Physiological and molecular determinants of embryo implantation.

In mammals, a new life begins with the union of an egg with a sperm, a process known as fertilization ( Wassarman, 1999 ). Following fertilization, the zygote undergoes several rounds of divisions and morphogenesis to form the blastocyst, an embryonic stage with two distinct cell lineages: the outer specialized trophectodermal epithelium and the inner cell mass ( Cockburn and Rossant, 2010 ; Wang and Dey, 2006 ). The blastocyst participates in the first physical and physiological interaction with the maternal endometrium to initiate implantation ( Red-Horse et al., 2004 ; Wang and Dey, 2006 ). A bidirectional crosstalk is essential for normal implantation thus the success of pregnancy, since perturbations will generate adverse outcomes for subsequent development, including decidualization and placentation, with potential loss of the pregnancy ( Chen et al., 2011 ; Song et al., 2002 ; Wilcox et al., 1999 ; Ye et al., 2005 ). Early pregnancy loss, occurring during the periimplantation period before pregnancy is recognized clinically, is a relatively common phenomenon in humans ( Cockburn and Rossant, 2010 ; Norwitz et al., 2001 ). For example, even in natural conception, the maximum chance of successful pregnancy occurring in a given menstrual cycle is limited to about 30% ( Zinaman et al., 1996 ). Only 50 to 60 percent of all conceptions advance beyond 20 weeks of gestation ( Norwitz et al., 2001 ). Among the pregnancies that are lost, implantation failure is the major cause, reaching approximate 75% ( Wilcox et al., 1988 ). Furthermore, 1 out of 7 couples worldwide are suffering from infertility ( Forti and Krausz, 1998 ). Despite significant developments in in vitro fertilization and embryo transfer (IVF-ET) technology that have overcome many underlying causes of infertility, pregnancy success rates remain relatively low, mainly due to implantation failure ( Miller et al., 2012 ; Norwitz et al., 2001 ; Wilcox et al., 1993 ). Therefore, it is imperative to address this global issue by investigating the mysteries of embryo implantation. Successful implantation requires synchronization between the acquisition of implantation competency by the blastocyst and a receptive state in the uterine endometrium ( Dey et al., 2004 ; Tranguch et al., 2005b ; Wang and Dey, 2006 ). These two events are precisely regulated by maternal hormones, in particular, ovarian estrogen and progesterone ( Conneely et al., 2002 ; Curtis Hewitt et al., 2002 ). Molecular and genetic evidence indicates that ovarian hormones together with locally produced signaling molecules, including cytokines, growth factors, homeobox transcription factors, lipid mediators and morphogen genes, function through autocrine, paracrine and juxtacrine interactions to specify the complex process of implantation ( Dey et al., 2004 ). However, the hierarchical landscape of the molecular signaling pathways that govern embryo-uterine interactions during early pregnancy remains to be explored in depth. The crosstalk between the blastocyst and the uterus can only occur during a brief period, namely the “window of implantation” ( Ma et al., 2003 ; Paria et al., 1993 ; Rogers and Murphy, 1989 ; Yoshinaga, 1980 ). In response to the implanting embryo, the surrounding uterine stroma undergoes cellular transformation, a process known as decidualization, to accommodate embryonic growth and invasion ( Lim and Wang, 2010 ). Locally induced decidua provides a positive feedback to support embryo survival. It is also thought that the decidua functions as a barrier against maternal immunological responses to the semi-allogenic embryo. However, it remains largely unclear how the blastocyst escapes maternal immune surveillance at the time of implantation. With the emergence of advanced technologies, a global analysis of gene and protein expression in the implanting embryo and uterus has been undertaken in several studies to unravel the molecular networks that control implantation in mice, as well as in humans ( Hamatani et al., 2004b ; Haouzi et al., 2011 ; Hu et al., 2008 ; Kao et al., 2002 ; Reese et al., 2001 ; Riesewijk et al., 2003 ; Yoon et al., 2004 ; Yoshioka et al., 2000 ). However, due to experimental difficulties and ethical restrictions, our understanding of human implantation still relies predominantly on animal models, particularly the mouse. Gene-knockout mouse models provide valuable information that has been used to construct a tentative molecular basis of implantation. Since embryo implantation is a dynamic developmental process that integrates many signaling molecules into a precisely orchestrated program, it is important to understand the hierarchical landscape of the pathways governing these processes to generate new strategies to correct implantation failure and improve pregnancy rates in women. This review will examine our understanding of signaling cascades that regulate embryo implantation and decidualization derived from gene expression studies and genetically engineered mouse models.

In the majority of eutherian mammals, implantation occurs in a fixed interval of time after ovulation when the corpus luteum is fully formed ( Finn and Martin, 1974 ). In humans, this is during the luteal phase of the menstrual cycle, while in rodents, it is in the diestrous phase of the estrous cycle. It has been well established that estrogen and progesterone are principal hormones in this process. According to their dynamic fluctuating levels, the reproductive cycle is divided into three stages ( Finn and Martin, 1974 ; Wang and Dey, 2006 ). The first stage is the proestrous or follicular phase in women during which estrogen levels are very high ( Michael, 1976 ; Yoshinaga et al., 1969 ). The second stage is a period when the levels of both hormones are low immediately after ovulation. Finally, the luteal stage is when both progesterone and estrogen are secreted from the corpus luteum. Embryo implantation occurs towards the end of the luteal phase. For example, at this stage in mice, the level of progesterone is gradually increased, owing to an enhanced secretion from newly formed corpora luteum, accompanied by a preimplantation surge of estrogen on day 4 of pregnancy (day 1=day of vaginal plug), while embryo implantation takes place at the midnight of day 4 ( McCormack and Greenwald, 1974 ; Wang and Dey, 2006 ) ( Figure 1A ). Based on the preimplantation ovarian steroid profiles, priming with exogenous estrogen and progesterone can confer on the uterus of ovariectomized mice a receptive state ( Lim et al., 1997 ; Paria et al., 1999b ). These hormones direct the preparation of the uterus for implantation in mice and rats ( Harper and Walpole, 1967 ; Roblero et al., 1987 ; Vinijsanun and Martin, 1990 ). It is generally accpted that progesterone is required for implantation nearly in all the animals studied, while the role of two estrogen surges at the proestrous and luteal phase prior to embryo implantation remains controversial ( Dey et al., 2004 ; Finn and Martin, 1972 ; Tranguch et al., 2006 ; Wang and Dey, 2006 ). It was once thought that progesterone alone mediates implantation. However, the observation that suckling mice and rats with normal progesterone secretion have a facultative delayed implantation suggests that there may be another hormone involved in implantation ( Malafaya et al., 2008 ; Mantalenakis and Ketchel, 1966 ; McLaren, 1968 ; Whitten, 1955 ; Yoshinaga and Adams, 1966 ). This suspicion is supported by the evidence that implantation can be induced in lactating rats by injection of a small dose of estrogen ( Krehbiel, 1941 ). Later studies provide direct evidence for the role of luteal estrogen in normal implantation ( Cochrane and Meyer, 1957 ; Whitten, 1958 ), showing that the timing of ovariectomy before or after luteal phase estrogen is critical for the induction of delayed implantation. For example, the blastocyst implants normally when ovariectomy is performed after preimplantation ovarian estrogen secretion, whereas if ovariectomy takes place before estrogen secretion, the embryo does not implant and the uterus enters into a condition of delayed implantation. With progesterone supplementation, the blastocyst remains quiescent, but it can be induced to implant by exogenous estrogen ( Paria et al., 1993 ). These findings clearly indicate that preimplantation estrogen secretion is crucial for blastocyst implantation into a progesterone-primed receptive uterus. Since preovulatory estrogen is secreted in most species, it is thought that proestrous estrogen optimizes the subsequent response just before implantation ( Barkley et al., 1979 ). Notably, the requirement for ovarian estrogen in implantation is species-specific. In species such as guinea pig, rhesus monkey, rabbit and golden hamster, progesterone alone is adequate for implantation ( Harper et al., 1969 ; Heap and Deanesly, 1967 ; Heap et al., 1981 ; Kwun and Emmens, 1974 ; Psychoyos, 1973 , 1986 ). However, participation of estrogen in implantation of these species may not be completely excluded. A hypothesis has been proposed that blastocysts in these species could synthesize and secrete estrogen locally to initiate implantation ( Dey et al., 2004 ; Wang and Dey, 2006 ). In agreement with this, aromatase, an enzyme for estrogen synthesis, is detected in the blastocyst of hamster and rabbit, while such an aromatase is absent in mice ( Dickmann et al., 1975 ; Hoversland et al., 1982a ; Reese et al., 2008 ; Sengupta et al., 1983 ; Sholl et al., 1983 ). It remains unclear whether blastocyst-uterus attachment during implantation requires ovarian estrogen in humans. In recent years, genetically engineered mouse models have provided valuable clues to understand the roles of progesterone and estrogen during embryo implantation. The function of progesterone is mainly mediated by its receptor, PR (encoded by Pgr gene), which has two isoforms, PRA and PRB ( Edwards, 2005 ). Both isoforms are expressed in the uterus ( Mote et al., 2006 ). Female mice lacking both PRA and PRB are infertile with many defects in ovarian and uterine functions ( Lydon et al., 1995 ), while these functions are normal in PRB deficient females ( Mulac-Jericevic et al., 2000 ), indicating that essential progesterone-regulated functions in uteri are primarily mediated by PRA. Estrogen functions in the uterus primarily through nuclear estrogen receptors ( Tan et al., 1999 ). ER also has two isoforms, known as ERα (encoded by Esr1 gene) and ERβ (encoded by Esr2 gene) ( Krege et al., 1998 ). Previous studies using knockout mice for ERs have demonstrated their differential functions in uterine biology ( Hewitt et al., 2005 ; Lee et al., 2012a ; Lubahn et al., 1993 ). ERα is the most important mediator of estrogen signaling during early pregnancy since ERα knockout mice are unable to support implantation ( Hewitt et al., 2005 ; Krege et al., 1998 ; Lee et al., 2012a ; Lubahn et al., 1993 ). Although ERβ knockout mice are fertile with normal implantation, increasing evidences show that it is also important in uterine biology ( Krege et al., 1998 ; Lee et al., 2012a ; Su et al., 2012 ; Wada-Hiraike et al., 2006 ). For example, ERβ is expressed in the endometrial endothelium and may participate in implantation through regulating angiogenic and vasomotor changes ( Huang et al., 2010 ; Su et al., 2012 ). In agreement with this, the expression level of ERβ is significantly lower in women with infertility ( Altmae et al., 2010 ). In addition, ERβ is also believed to be a potent player in human labor onset due to its high expression in the myometrium and the cervix ( Huang et al., 2010 ; Su et al., 2012 ).

Acquisition of implantation competency by the blastocyst is a prerequisite for successful implantation ( Paria et al., 1993 ). In mice, the blastocyst escapes from its zona pellucida and attaches to the uterine epithelium at day 4.5 of pregnancy ( Das et al., 1994 ; McCormack and Greenwald, 1974 ; Wang and Dey, 2006 ). However, this sequence is interrupted in delayed implantation. Except for the facultative delay in lactating mice and rats ( Enzmann et al., 1932 ; Hamlett, 1935 ; Kirkham, 1918 ), implantation delay occurs naturally as an obligate delay in nearly 100 mammalian species, notably the mustelids and marsupials ( Lopes et al., 2004 ; Mead, 1993 ; Renfree and Shaw, 2000 ; Thom et al., 2004 ). But in some species such as the hamster, guinea pig, rabbit and pig, delayed implantation does not occur ( Dey et al., 2004 ). It is unclear whether this phenomenon exists in humans. Delayed implantation can be induced experimentally in mice and rats by ovariectomy before the preimplantation ovarian estrogen surge and maintained by injection of progesterone ( Huet and Dey, 1987 ; Yoshinaga and Adams, 1966 ), providing a powerful model for the study of implantation. During delayed implantation, the blastocyst is metabolically dormant and incompetent to initiate attachment in the uterus ( Nieder and Weitlauf, 1985 ; Van Blerkom et al., 1978 ). Although embryos develop into blastocysts and undergo zona dissolution, they show signs of being inactive with subnormal metabolic activity, reduced cell divisions with low DNA synthesis ( Lopes et al., 2004 ). Ultrastructural observations of the trophoblast cells in dormant blastocysts reveal several morphological fine structure changes. For example, the ribosomes become monosomes, the endoplasmic reticulum is less profiled and the Golgi apparatus is not well developed ( Renfree and Shaw, 2000 ; Wu and Meyer, 1974 ). The blastocyst can maintain this dormant state within the uterine cavity for days or even weeks ( Lee et al., 2011a ). But under appropriate conditions, the blastocyst is rapidly activated to resume its development with attachment and invasion. The active blastocyst also shows distinct morphological differences from the dormant blastocyst. For example, a more irregular surface with more microvilli is observed in activated trophoblast cells, with accumulating glycogen granules in the cytoplasm ( Naeslund et al., 1980 ). Using the murine delayed implantation model and blastocyst transfer techniques, it has been demonstrated that dormant blastocysts fail to implant in the receptive uterus, highlighting the notion that the blastocyst’s state of activity determines the “window” of implantation in mice ( Paria et al., 1993 ). However, it is still largely unknown how blastocysts undergo dormancy and survive for an extended period or how they become reactivated to acquire implantation competency. Since implantation is a two-way interaction, it has been speculated that there is a growth-impeding substance in the uterine secretions that target the embryos ( Surani, 1975 ). In support of this hypothesis, a blastocyst in delay that is subsequently activated reverts to its delayed state if replaced in the uterine cavity of a mouse in delayed implantation ( Naeslund et al., 1980 ). Another explanation of embryo dormancy or diapause is that suboptimal conditions exist for blastocyst growth. Several in vitro experiments support this view. For instance, depletion of glucose and/or arginine and leucine from a conventional medium impedes blastocyst growth ( Nieder and Weitlauf, 1985 ). Although these observations are valuable for the clues about blastocyst activation, the underlying molecular and cellular mechanisms are still unknown. The advent of cDNA microarray technology and genomic sequencing approaches has made global analysis of differential gene expression between dormant and active blastocysts possible ( Hamatani et al., 2004b ). Hamatani and coworkers find that 229 genes are differentially expressed between dormant and activated blastocysts, suggesting that the physiological states are molecularly distinguishable. The major functional categories of altered genes include the cell cycle, cell signaling, and energy metabolic pathways. Current knowledge on mechanisms governing the process of blastocyst activation for implantation is summarized in Figure 2 . The dormant blastocysts in delayed implantation in mice will resume implantation upon an injection of estrogen, suggesting the importance of estrogen in mediating both the process of uterine receptivity and blastocyst activation ( Paria et al., 1993 ). Since estrogen receptor is present in preimplantation mouse embryos ( Hou and Gorski, 1993 ; Hou et al., 1996 ), it was previously thought that estrogen could trigger blastocyst activation for implantation. However, a specific ER antagonist can not inhibit the active status of the blastocyst, and estrogen fails to activate dormant blastocysts in culture, suggesting the participation of a pathway distinct from the classical nuclear ERs signaling ( Paria et al., 1998 ). A further study shows that 4-hydroxyestradiol (4-OH-E2), a catechol metabolite of endogenous estrogen, is effective in initiating blastocyst activation through stimulating the synthesis of prostaglandin (PG) ( Paria et al., 1998 ). A haloestrogen, 2-fluoroestrdiol-17β (2-FL-E2), is a potent estrogen but acts as an inhibitor of catecholestrogen synthesis. It has been previously shown to stimulate uterine growth and estrogen-targeting gene expression in mice ( Dey et al., 1986 ; Mitchell et al., 1993 ; Paria et al., 1990 ; Paria et al., 1998 ). To distinguish the site-specific function of estrogen and catecholestrogen in uterine preparation versus blastocyst activation for implantation, experiments employing 2-FL-E2 have demonstrated that, unlike native estrogen, 2-FL-E2 fails to induce implantation in progesterone-primed delayed implantation mice, even at a relatively high dose ( Paria et al., 1998 ). Moreover, dormant blastocysts fail to implant after their transfer into uteri of progesterone-treated delayed recipients receiving an injection of 2-FL-E2, whereas normal day 4 blastocysts show implantation after transfer into recipient receiving the same treatment ( Paria et al., 1998 ). These results indicate that 2-FL-E2 can induce the uterus into a receptive status, but fails to activate the dormant blastocyst mainly due to its inability to transform into catecholestrogen. In this respect, catecholestrogen alone can re-induce blastocyst activation and attachment reaction in delayed implanting mice ( Hoversland et al., 1982b ; Kantor et al., 1985 ). Along with this finding, the receptive uterus is capable of transforming the native estrogen into catecholestrogen prior to embryo implantation ( Paria et al., 1990 ; Paria et al., 1998 ). CYP1B1, an enzyme involved in NADPH-dependent 4-hydroxylation of estrogens, is expressed in uterine stroma cells on day 4 of pregnancy in mice ( Paria et al., 1998 ; Reese et al., 2001 ). Observations of the presence of CYP1B1 activity in the progesterone-treated uterus and Cyp1b1 transcripts in the receptive uterus strongly suggest that catecholestrogen locally produced in the uterus directs blastocyst activation for implantation. Collectively, ovarian estrogen interacting via the nuclear ERs is required for the preparation of the receptive uterus, whereas its catechol metabolite 4-hydroxyestradiol produced locally in the uterus activates the blastocyst for implantation. Future studies are warranted to further reveal the molecular mechanisms by which catecholestrogen exerts bioactivity and whether there is a specific receptor solely for catecholestrogen in the blastocyst and the uterus during implantation. It is also equally important to address whether and how ovarian steroid hormones and their metabolites interact with other local factors to initiate embryo implantation during early pregnancy. Psychoactive cannabinoids are major components in marijuana and exert most of their effects through activation of G protein-coupled cell surface receptors, cannabinoid receptor 1 (CB1) and cannabinoid receptor 2 (CB2), encoded in mice by the genes Cnr1 and Cnr2 , respectively ( Wang et al., 2006a ). With the discovery of cannabinoid receptors, two endogenous cannabinoid ligands, arachidonoylethanolamide (also known as anandamide) and 2-arachidonolglycerol, have been identified that can be synthesized in the uterus during early pregnancy ( De Petrocellis et al., 2004 ; Maccarrone and Finazzi-Agro, 2002 ; Paria et al., 1996 ). The receptor-mediated activity of anandamide depends on its extracellular concentration, which largely depends on its intracellular degradation by the endocannabinoid-degrading enzyme, fatty acid amide hydrolase (FAAH) ( Cravatt et al., 2001 ; Ueda et al., 2000 ; Wang et al., 2006c ). Endocannabinoid signaling has recently been highlighted as an important lipid-mediated pathway that directs preimplantation embryo development and the timely homing of embryos into the uterus ( Paria et al., 2001b ; Wang and Dey, 2005 ; Wang et al., 2004a ; Wang et al., 2006b ). Although both CB1 and CB2 are expressed in preimplantation embryos ( Paria et al., 1995 ), cellular and pharmological experiments have demonstrated that CB1 is the functional receptor subtype in the preimplantation embryos. For example, two-cell embryos cultured in the presence of natural, synthetic or endocannabinoids fail to develop into the blastocyst stages, while a CB1-selective antagonist can reverse this inhibitory effect ( Schmid et al., 1997 ; Wang et al., 1999 ). These findings indicate that the mouse embryo is a potential target for cannabinoid signaling. During the periimplantation period, the levels of uterine anandamide and blastocyst CB1 receptor are precisely regulated. Both the ligand and CB1 receptor are maintained at high levels in the nonreceptive uterus and dormant blastocysts, whereas with the attainment of uterine receptivity and blastocyst activation, they are coordinately downregulated ( Das et al., 1995 ; Guo et al., 2005 ; Schmid et al., 1997 ). This is further supported by the finding showing a higher level of anandamide in leukemia inhibitory factor deficient ( Lif -/- ) mice with implantation failure than those in wild-type mice ( Paria et al., 2001b ). Similar fluctuations in anandamide levels during early pregnancy have been observed in women undergoing IVF-ET treatment, showing that a significantly lower level of anandamide during implantation is required for successful pregnancy ( El-Talatini et al., 2009 ). Indeed, a very narrow range of anandamide regulates blastocyst function by differentially modulating mitogen-activated protein kinase (MAPK) signaling and Ca 2+ -channel activity via CB1 ( Wang et al., 2003 ). For example, anandamide at a low concentration induces the activation of MAPK signaling and confers implantation competency on the blastocyst, whereas a higher concentration dampens this response by inhibiting Ca 2+ mobilization ( Wang et al., 2003 ). Thus, it is conceivable that critical levels of endocannabinoids produced in the uterus synchronize blastocyst activation with uterine receptivity for implantation. Aberrant levels of uterine endocannabinoids or blastocyst CB1 receptors interfere with the normal embryo-uterine dialogue during implantation, resulting in early pregnancy loss. It is noteworthy that spontaneous pregnancy losses are associated with elevated anandamide levels in women ( Habayeb et al., 2008 ; Maccarrone et al., 2002 ; Maccarrone et al., 2000 ), reinforcing the concept that endocannabinoid signaling is an important determinant of embryonic fate during implantation. The highly conserved Wnt signaling pathway is a critical mediator of cell-cell interactions during embryogenesis ( van Amerongen and Nusse, 2009 ), it is thus assumed to be involved in the process of implantation, referring to intricate interplay between the blastocyst and the uterus. Wnt signaling is branched into canonical and noncanonical pathways according to its distinct functions ( van Amerongen and Nusse, 2009 ). With respect to canonical Wnt pathway, the Wnt ligands transduce their signals through combining with the frizzled (Fzd) receptor and low density lipoprotein receptor-related protein (LRP) Lrp5/6 co-receptors complex ( Logan and Nusse, 2004 ). This binding will liberate the cytoplasmic β-catenin, which is under continuous proteasome-mediated degradation before the pathway is activated ( Metcalfe and Bienz, 2011 ). Then intracelluar accumulated β-catenin will translocate into the nucleus and form a complex with the lymphoid enhancer-binding factor/T cell-specific transcription factor (LEF/TCF) to guide the transcription of target genes ( Behrens et al., 1996 ). The non-canonical pathway is independent of β-catenin via solely binding to Fzd receptors, and the downstream effectors are diverse, including Ca 2+ /Planar Cell Polarity (PCP) pathway, Rho signaling, Wnt-PKA signaling and so on ( Barrow, 2006 ; Semenov et al., 2007 ). The secreted frizzled related proteins (sFRPs), which have similar sequence signatures to Fzd receptors in the cysteine-rich domain to bind the ligand but lack the transmembrane domain to conduct the signal, will inhibit the bioactivities of Wnt proteins ( Rattner et al., 1997 ; Suzuki et al., 2004 ). Alternatively, the co-receptor Lrp5/6 can also be regulated by the Dickkopf proteins (Dkks) to downregulate cell-surface LRP receptors through interacting with Kremen ( Bafico et al., 2001 ; Glinka et al., 1998 ; Mao and Niehrs, 2003 ; Mao et al., 2002 ; Mao et al., 2001 ; Semenov et al., 2001 ). The complexity and redundancy of the Wnt family of proteins, receptors, extracellular antagonists and intracellular signaling components suggest that the nature of the inter- and intra-cellular Wnt machinery determines the orchestration of this signal pathway during development. Previous studies have revealed specific expression pattern of Wnt pathway members in early embryos and uteri during the periimplantation period in mice ( Hamatani et al., 2004a ; Kemp et al., 2005 ; Mohamed et al., 2004 ; Mohamed et al., 2005 ; Paria et al., 2001a ; Wang et al., 2004c ; Zeng et al., 2004 ), and the dispensable role of canonical Wnt signaling in blastocyst formation is well studied and reviewed previously ( Chen et al., 2009 ; Sonderegger et al., 2010 ). Notably, during implantation, the canonical Wnt signaling is required for blastocyst competency. Female mice with oocyte specific deletion of β-catenin give reduced number of pups compared with the wild-type littermates, as embryos are lost during the blastocyst stage, implying the role of β-catenin-dependent pathway during the periimplantation stage ( De Vries et al., 2004 ). Interfering the Wnt pathway with the sFRP molecule also disturbs the implantation process ( Mohamed et al., 2005 ). Using the strategy of adenoviral mediated Dkk1, our previous study has clearly demonstrated that silencing canonical Wnt/β-catenin signaling does not adversely affect the uterine preparation for receptivity, but remarkably blocks blastocyst competency for implantation in mice ( Xie et al., 2008 ). The significance of this pathway in blastocysts is first evidenced from the findings utilizing delayed implantation model, showing that the activity of nuclear β-catenin signaling is different between the dormant and reactivated blastocysts ( Xie et al., 2008 ). Moreover, Wnt3a is able to induce intracellular accumulation and nuclear translocation of β-catenin in trophectoderm cells and concurrently induces the expression of peroxisome proliferator-activated receptor (PPAR) δ, a nuclear receptor for prostacyclin ( Xie et al., 2008 ). Through pharmological gain of function study we further reveals that canonical Wnt signaling synergizes with PG signaling to confer blastocyst competency to implantation ( Xie et al., 2008 ). This is consistent with the early finding that expression of cyclooxygenase 2 (COX-2), a rate-limiting enzyme of PG biosynthesis, is substantially upregulated in activated blastocysts after treatment with catecholestrogen ( Paria et al., 1998 ). Collectively, these findings constitute direct evidence that Wnt signaling is at least one pathway determining blastocyst competency for implantation. Although it is clear that a maternal-embryonic dialogue is essential for the initiation of embryo implantation and that the state of the blastocyst determines the implantation window ( Paria et al., 1993 ), there is a long-standing quest for specific embryo-derived signaling molecules that can functionally influence the uterus, as well as other reproductive organs ( Lim and Dey, 2009 ). Global gene analysis of the dormant versus active blastocysts demonstrates that heparin-binding epidermal growth factor (EGF)-like growth factor (HB-EGF) encoded by Hbegf gene is significantly up-regulated during blastocyst activation ( Hamatani et al., 2004b ). Prior studies demonstrate that Hbegf is expressed in the luminal epithelium at the site of blastocyst apposition approximate 6 hours prior to the blastocyst attachment reaction in mice ( Das et al., 1994 ; Paria et al., 1999a ). Moreover, increasing expression of its receptors (ErbB1 and ErbB4) and ligand binding activity has been observed in the blastocysts that are competent for implantation ( Das et al., 1997 ; Paria et al., 1993 ; Raab et al., 1996 ). Most interestingly, blastocyst-size Affi-gel beads presoaked with HB-EGF protein that are transferred intraluminally into a pseudopregnant uterus can induce its own gene expression in uterine cells surrounding the beads and increase vascular permeability, similar to the physiological changes induced by normal blastocysts ( Hamatani et al., 2004b ; Paria et al., 2001a ). Signaling by HB-EGF back to the embryo, in turn, activates the program of trophoblast differentiation required for adhesive functions during subsequent attachment and invasion ( Wang et al., 2000 ). These observations suggest that HB-EGF conducts an auto-induction loop between the implanting blastocyst and the uterus via paracrine and juxtacrine manners. Maternal deficiency of HB-EGF targeted to the uterus defers the window of implantation and compromises pregnancy outcome, while amphiregulin, another heparin-binding member of the EGF family, partially compensates for the loss of HB-EGF during implantation ( Xie et al., 2007 ). In human, HB-EGF is highly expressed in the receptive endometrium ( Birdsall et al., 1996 ; Leach et al., 1999 ; Stavreus-Evers et al., 2002 ; Yoo et al., 1997 ) and its receptor ErbB4 is localized on the surface of the trophectoderm in periimplantation blastocysts ( Chobotova et al., 2002b ), indicating that HB-EGF-ErbB4 signaling also mediates the trophectoderm-uterine epithelium interaction during implantation in humans. HB-EGF continues to regulate trophoblast development and survival during placentation in the first trimester of human gestation ( Jessmon et al., 2009 ). As described above, aberrant endocannabinoid signaling is detrimental for preimplantation development and blastocyst implantation ( Paria et al., 2001b ; Wang et al., 1999 ). As an endogenous cannabinoid ligand, anandamide is spatiotemporally regulated in the uterus during early pregnancy, i.e., its expression levels are lower in the receptive uterus than for non-receptive uterus and at the implantation site than for interimplantation site ( Paria et al., 2001b ; Wang et al., 1999 ). Regarding the machinery controlling appropriate levels of anandamide in the implanting uterus, it is worth noting that the blastocyst can rapidly release a non-identified, cell permeable lipid that is able to activate uterine FAAH, protecting against the detrimental effects of excessive uterine endocannabinoids ( Maccarrone et al., 2004 ). Another intriguing example of embryonic regulation of uterine function during implantation is revealed by asynchronous embryo transfer studies. When preimplantation embryos at different developmental stage are transferred into the oviducts of the same recipient, embryos of advanced stage implant in advance of younger embryos prior to the anticipated time when implantation normally occurs ( Ueda et al., 2003 ). This finding indicates that the presence of embryos in the reproductive tract can prime the endometrium through unknown embryonic factors, leading to an expansion of the implantation window. A previous study employing non-human primates reports that super-imposition of endometrial receptivity on embryonic stimuli causes remarkable changes in endometrial expression of pro-inflammatory cytokines and immunosuppressive factors, which regulate endometrial transformation conducive for embryo survival, growth and development ( Nimbkar-Joshi et al., 2009 ). A similar phenomenon has been observed in Lif null pregnant mice, in which implantation fails to occur, but the uterine luminal epithelium exhibits COX-2 expression at the apposition site of blastocysts ( Fouladi-Nashta et al., 2005 ; Song et al., 2000 ), reinforcing the notion that embryo-derived signaling molecules are able to regulate uterine preparation for implantation. It is a challenge to unravel the nature of embryonic signals influencing uterine functions. One of the limiting factors is the availability of adequate amount of tissues for analysis. With the advent of microscale proteomics and genomics, it becomes possible to identify embryonic signals during implantation ( Aghajanova et al., 2012 ; Dominguez et al., 2010 ; Katz-Jaffe et al., 2006 ; Tang et al., 2009 ). Identification of putative secreted embryonic signaling molecules could be validated by employing the uterine blastocyst-size bead transfer approach, as previously described ( Hamatani et al., 2004b ; Paria et al., 2001a ). After the attachment, the embryo directly contacts with the endometrium and the uterine stroma cells initiate the decidualization that embed and protect the embryo. Decidualization can also occur in response to artificial stimulus by oil injection or endometrium scratch in the receptive uterus to form deciduoma, which exclude the absolute requirement of the conceptus ( Loeb, 1906 , 1908 ). However, some studies uncover the morphological and time schedule difference between the deciduoma and the decidua, which hint the involvement of embryo-derived signals ( Deanesly, 1971 ; Lundkvist and Nilsson, 1982 ; Welsh and Enders, 1985 ). The disappearance of primary decidual zone which forms a barrier surrounding the embryo and reduced immune cell population in deciduoma further emphasize the importance of embryo-uterus dialogue ( Herington and Bany, 2007b ; Wang et al., 2004d ). Using microarray approaches, the differentially expressed genes responsive to ( Bany and Cross, 2006 ; Herington and Bany, 2007a ; Herington and Bany, 2007b , 2009 ; Herington et al., 2009 ; Kashiwagi et al., 2007 ; McConaha et al., 2011 )the living embryo versus artificial stimuli further prove the paracrine signals from the embryo regulating the decidua development ( Bany and Cross, 2006 ; Herington and Bany, 2007a ; Herington and Bany, 2007b , 2009 ; Herington et al., 2009 ; Kashiwagi et al., 2007 ; McConaha et al., 2011 ). Collectively, these findings highlight the necessity of embryo-derived signals for normal implantation and decidualization.

In placental mammals, the uterus is receptive to blastocyst implantation during a spatiotemporally restricted “window” when the uterine environment is favorable for blastocyst implantation ( Yoshinaga, 1988 ). In mice, this period is limited to day 4 of pregnancy ( Wang and Dey, 2006 ). The uterus cannot initiate implantation before this period (days 1–3). Immediately after the receptive state, the uterus spontaneously enters a refractory phase (day 5 onward) where the uterine environment is hostile for blastocyst survival ( Carson et al., 2000 ; Yoshinaga, 1988 ). Similarly, in humans, the receptivity period occurs between days 20 and 24 of a regular menstrual cycle (days 6 to 10 after ovulation LH surge) ( Bergh and Navot, 1992 ; Lessey, 2011 ; Noyes et al., 1975 ; Psychoyos, 1973 ; Rashid et al., 2011 ). The key to uterine receptivity is the dynamic and precisely controlled molecular and cellular events that drive blastocyst growth, attachment and the subsequent events of implantation. This dynamic process involves a variety of genes that include cytokines, homeobox transcription factors, and developmental genes working together with ovarian steroids to specify uterine receptivity. Ovarian steroids progesterone and estrogen are principal hormones that direct uterine receptivity. Synchronized production of progesterone and estrogen mediates structural and functional changes in the uterus that enable the blastocyst to attach and initiate the process of implantation. On day 1 of pregnancy in mice, under the influence of preovulatory ovarian estrogen, the uterine epithelial cells undergo extensive proliferation that to some extent continues through day 2. Rising progesterone levels secreted from the newly formed corpora luteum initiate stromal cell proliferation from day 3 onward ( Huet-Hudson et al., 1989 ; Huet et al., 1989 ). On the morning of day 4, when the uterus enters the pre-receptive stage, the production of a small amount of estrogen is crucial for the uterus to attain receptivity ( Tranguch et al., 2005b ). At that time, the uterine epithelial cells gradually lose their polarity, and meanwhile, the plasma membranes of the epithelial cells become smooth and flattened at the site of blastocyst apposition ( Martin et al., 1970 ). To some extent, the window of uterine receptivity is flexible and can be modified under different hormonal environments. In mice, the blastocyst can initiate implantation outside of the normal “window” of uterine receptivity ( Song et al., 2007 ). For example, blastocysts can initiate the attachment reaction in the non-receptive uterus when transferred on day 5 of pseudopregnancy, but implantation will not occur when normal blastocysts are transferred into the day 6 pseudopregnant uterus ( Song et al., 2007 ). Exogenous progesterone supplementation can prolong the implantation window to day 6 with sustained LIF expression ( Song et al., 2007 ). However, this deferred embryo implantation leads to embryonic demise before birth in mice, and is often associated with higher risk of early pregnancy losses in humans ( Wilcox et al., 1999 ). Estrogen is a critical determinant specifying the duration of the window of uterine receptivity for implantation. Using different doses of estrogens in the delayed implantation model, estrogen extends the window of uterine receptivity at a low threshold, whereas physiologically higher levels rapidly close the implantation window, transforming the uterus into a refractory state ( Ma et al., 2003 ). Although whether blastocyst attachment reaction requires ovarian estrogen or not is still ambiguous in humans, exposure to high levels of estrogen, which could be resulting from ovarian stimulation with clomiphene citrate in IVF, leads to implantation failure and embryo resorption ( Ertzeid and Storeng, 2001 ; Shapiro et al., 2011 ). This reduced implantation rate in IVF cycles could be due to asynchrony between the endometrium and the blastocyst when exposed to high levels of estrogen ( Devroey et al., 2004 ). Of the cytokines that have been studied, LIF is most pertinent to implantation ( Kimber, 2005 ; Stewart et al., 1992 ; White et al., 2007 ). The LIF receptor shares gp130 as a common signal-transduction partner with other cytokine receptors. Expression of LIF is biphasic on day 4, initially in the uterine glands and later in the stromal cells surrounding the blastocyst during the attachment reaction ( Ni et al., 2002 ; Song and Lim, 2006 ; Song et al., 2000 ). This expression pattern indicates that LIF has dual roles, first in uterine preparation and later in the attachment reaction ( Song et al., 2000 ; Stewart et al., 1992 ). Lif -deficient female mice exhibit implantation failure and supplementation with LIF rescues this defect ( Chen et al., 2000 ; Stewart et al., 1992 ). Moreover, pharmological blocking LIF action in the uterus significantly reduces the phosphorylation of the downstream signaling molecule signal transducer and activator of transcription (STAT) 3 in the uterus, resulting in implantation failure ( Menkhorst et al., 2011 ; Mohamet et al., 2009 ; White et al., 2007 ). The importance of LIF signaling in implantation is further evidenced by the observations that inactivation of gp130 and STAT3 also causes implantation failure ( Catalano et al., 2005 ; Cheng et al., 2001 ; Daikoku et al., 2011 ; Ernst et al., 2001 ). In humans, LIF is expressed at high levels in the glandular epithelium of the secretory endometrium ( Rashid et al., 2011 ), with a significant increase in the luminal and glandular epithelium at mid-secretory phase ( Leach et al., 2012 ). It is reported that an optimal level of LIF is required for blastocyst implantation in women ( Menkhorst et al., 2011 ; Terakawa et al., 2011 ). Moreover, clinical evidence shows that LIF deficiency is associated with unexplained recurrent abortion and infertility in women ( Dey et al., 2004 ; Ernst et al., 2001 ; Hambartsoumian, 1998 ). Overall, significant gains have been made in our understanding of the function of LIF, a critical determinant of embryo implantation. Homeobox genes are evolutionarily conserved transcriptional regulators that control embryonic morphogenesis and differentiation ( Krumlauf, 1994 ). In mice, homeobox A (Hoxa) genes, Hoxa10 and Hoxa11 , are expressed in uterine stromal cells during receptivity and upregulated upon decidualization in a steroid hormone-responsive manner ( Benson et al., 1996 ; Gendron et al., 1997 ; Hsieh-Li et al., 1995 ). This specific expression pattern indicates that these two transcription factors may have dual roles, first in uterine receptivity and then in decidualization. Both Hoxa10 and Hoxa11 mutant mice are infertile due to implantation defects ( Bagot et al., 2001 ; Gendron et al., 1997 ; Lim et al., 1999b ). Hoxa11 -/- uteri have a more severe phenotype than Hoxa10 -/- mice, and the absence of Lif expression in Hoxa11 -/- uteri reinforces its role as a crucial participant in uterine receptivity and later events of implantation ( Gendron et al., 1997 ). Regarding the pathophysiological significance of Hox genes in human implantation, expression of HOXA10 and HOXA11 in the endometrium increases significantly in the midluteal phase when the uterus is receptive for embryo attachment ( Gui et al., 1999 ; Taylor et al., 1998 ; Taylor et al., 1999b ), and is significantly lower in infertile women ( Eun Kwon and Taylor, 2004 ; Fischer et al., 2011 ; Matsuzaki et al., 2009 ; Taylor et al., 1999a ). Nonclassical Hox genes may also be important in implantation. Ablation of H6 homeobox 3 ( Hmx3 ) in mice leads to implantation failure ( Wang et al., 1998 ). Msx1 , another homeobox gene, is transiently expressed in the mouse luminal epithelium and glandular epithelium on the morning of day 4 of pregnancy ( Daikoku et al., 2004 ; Pavlova et al., 1994 ), but its expression is dramatically downregulated to undetectable levels upon the termination of uterine receptivity ( Daikoku et al., 2004 ; Pavlova et al., 1994 ). However, in Lif -/- mice, Msx1 is consistently expressed in the uterine epithelium even on day 6 of pregnancy, suggesting that LIF signaling is essential for the down-regulation of Msx1 ( Daikoku et al., 2004 ). This is confirmed by the observation of sustained Msx1 expression in uteri with conditional depletion of gp130 ( Daikoku et al., 2011 ). Conditional deletion of uterine Msx1 impairs implantation. Histological analysis of Msx1 -/- implantation sites reveals that the luminal epithelium lacks well-defined crypts for blastocyst homing and attachment ( Daikoku et al., 2011 ). Moreover, a double knockout of uterine Msx1 and Msx2 results in complete implantation failure with altered luminal epithelial cell polarity and impaired stromal-epithelial communication ( Daikoku et al., 2011 ; Nallasamy et al., 2012 ), pointing toward a compensatory role of Msx2 in the establishment of uterine receptivity in the absence of Msx1 . Nonetheless, these results suggest that Msx1/2 genes are critical for conferring uterine epithelial integrity and uterine receptivity in mice. The dialogue between the blastocyst and uterus shares features with reciprocal epithelial-mesenchymal interactions that occur during embryogenesis, and involves evolutionarily conserved signaling pathways mediated by Indian hedgehog (IHH) proteins, bone morphogenetic proteins (BMPs), Wnts, and their cognate receptors ( Paria et al., 2001a ). IHH is a member of the hedgehog family which fine-tune cell proliferation, differentiation and cell-cell communication in many processes such as organogenesis, stem cell maintenance and oncogenesis ( Ingham and McMahon, 2001 ; Ryan and Chiang, 2012 ). The progesterone-dependent expression of Ihh is restricted to the epithelium, whereas the effectors of IHH, patched-1 ( ptch1 ) and transcription factors GLI-Krüppel family member ( Gli) 1–3 are coordinately expressed in the underlying stroma ( Matsumoto et al., 2002 ; Takamoto et al., 2002 ). Conditional deletion of Ihh in the uterus results in implantation failure due to a lack of progesterone-facilitated stromal proliferation, angiogenesis and decidualization, phenotypically mimicking that of PR knockout mice ( Franco et al., 2010a ; Lee et al., 2006 ). These findings indicate that IHH is an essential mediator for PR action in the uterus, participating in the communication between the uterine epithelium and the stroma required for embryo implantation. Microarray analysis demonstrates that IHH mRNA is synchronously regulated with PR during the human menstrual cycle ( Talbi et al., 2006 ). Indeed, the temporal elevation of endometrial IHH and GLI1 during the secretory phase, and their modulation by CDB-2914, a selective PR modulator ( Wei et al., 2010 ), suggest that progesterone regulated hedgehog signaling is a key regulator of human endometrial differentiation and implantation. Moreover, endometrial IHH expression is reduced in women with endometriosis, indicative of progesterone resistance ( Smith et al., 2011 ). BMPs are the largest family of morphogens belonging to the transforming growth factor-β (TGF-β) superfamily of growth modulators ( Zamani and Brown, 2011 ), and transcripts corresponding to several BMP family members are expressed in mouse uteri ( Paria et al., 2001a ; Ying and Zhao, 2000 ). Among all the BMPs expressed in the uterus, only BMP2 is induced in response to progesterone, with intense expression in the stromal cells surrounding the implanted embryo ( Paria et al., 2001a ). In vitro studies have demonstrated that addition of recombinant BMP2 to undifferentiated stromal cells markedly enhances the decidualization by stimulating Smad signaling pathway, whereas silencing the expression of BMP2 in these cells efficiently blocks the decidualization in both mice and humans ( Lee et al., 2007 ; Li et al., 2007 ). Moreover, the biological activity of BMP2 is mediated by proprotein convertase 6 (PC6) during decidualization; therefore, addition of recombinant active BMP2 partially rescues the decidualization arrest caused by PC6 inhibition ( Heng et al., 2010 ). This is consistent with the previous findings that PC6 is highly expressed and regulated during implantation and decidualization, and that deletion or knockdown of PC6 inhibits decidualization, causing implantation failure and female infertility ( Nie et al., 2005 ; Okada et al., 2005 ; Tang et al., 2005 ). Interestingly, uterine expression of the PR-interacting protein, Krüppel-like factor (KLF) 9, is significantly enhanced upon conditional ablation of Bmp2 ( Lee et al., 2007 ; Li et al., 2007 ). Previous studies have demonstrated that KLF9 is highly expressed in the predecidual stroma and becomes undetectable in the decidua, while its deficiency induces subfertility in mice and reduces uterine progesterone sensitivity ( Simmen et al., 2004 ; Velarde et al., 2005 ). Increasing evidence points toward the existence of a negative feedback loop between KLF9 and BMP2 to maintain PR function in stromal cells for successful implantation, where deregulation of this interaction can compromise the attainment of uterine receptivity and cause infertility ( Pabona et al., 2010 ). Gene expression profiling experiments have identified Wnt4 as a downstream target of BMP2-induced decidualization ( Li et al., 2007 ). Wnt4 is expressed primarily in the luminal epithelium during the preimplantation period and then re-localizes to the stromal cells surrounding the implanting embryo and expands its expression to the decidua ( Daikoku et al., 2004 ; Hayashi et al., 2009 ). Conditional deletion of Wnt4 in the uterus renders female mice subfertile due to defective embryo implantation and subsequent decidualization ( Franco et al., 2011a ). Molecular analysis demonstrates that Bmp 2 along with other decidual marker genes such as follistatin are downregulated in Wnt4 -null uteri ( Franco et al., 2011a ). These data indicate a complex correlation between Wnt4 and BMP2 during decidualization. In addition to Wnt4, many other components of Wnt signaling pathway are spatiotemporally regulated in the periimplantation uterus and are believed to be crucial to implantation ( Chen et al., 2009 ; Hayashi et al., 2009 ). For example, Wnt5a is expressed in the uterine subepithelial stroma albeit at lower levels in the epithelium ( Hayashi et al., 2009 ; Hou et al., 2004 ; Mericskay et al., 2004 ). However, its expression is upregulated in both the epithelium and stroma in Msx1 -/- and Msx1/Msx2 -/- uteri ( Daikoku et al., 2011 ; Nallasamy et al., 2012 ). In addition, application of Wnt5a in vitro compromises blastocyst invasion and trophoblast outgrowth when co-cultured with uterine epithelial cells ( Daikoku et al., 2011 ). These findings suggest that Wnt5a may affect epithelial function and, thus, direct blastocyst trophectoderm differentiation. A recent study show that Wnt7b is strongly upregulated in uterine epithelia after periimplantation estrogen secretion and that its expression can be induce by injection of estrogen in ovariectomized mice ( Hayashi et al., 2009 ). Therefore, it is likely that Wnt7b is a direct mediator of uterine function in response to ovarian estrogen. However, a definitive role of Wnt7b in uterine receptivity warrants further investigation through the use of a conditional deletion mouse model, since Wnt7b null mice die during midgestation due to a failure of chorioallantoic fusion ( Parr et al., 2001 ). Wnt7a is also expressed in the luminal epithelium with maximal expression on day 5 and declines thereafter at both implantation and inter-implantation sites ( Hayashi et al., 2009 ). In contrast to Wnt7b, uterine expression of Wnt7a is not under the control of ovarian estrogen. Adult Wnt7a -deficient mice are viable, but infertile, owing to a lack of endometrial glands, suggesting that Wnt7a is crucial for normal formation and maintenance of uterine cellular architecture ( Carta and Sassoon, 2004 ; Miller and Sassoon, 1998 ; Parr and McMahon, 1998 ). Sfrp4 , a Wnt inhibitory protein, is expressed in the uterine stroma during the receptive phase ( Daikoku et al., 2004 ). A marked downregulation of sfrp4 in Lif / uteri indicates that sFRP4 could participate in regulating uterine preparation for implantation ( Daikoku et al., 2004 ). This is further supported by the observation that sFPR4 expression remains low in the endometrium of women with recurrent implantation failure ( Revel et al., 2011 ). Unlike sFRP4, expression of sFRP2 is downregulated at the implantation site, and intraluminal infusion of sFRP2 protein disrupts normal implantation ( Mohamed et al., 2005 ). This differential expression of Wnts and their inhibitory molecules underscores that a precisely regulated Wnt system is crucial to uterine preparation for implantation.

Embryo implantation is a dynamic developmental event that involves a series of physical and physiological interactions among the blastocyst trophectoderm and various endometrial cell-types, including the luminal and glandular epithelial and stromal cells ( Aplin and Kimber, 2004 ; Carson, 2008 ; Enders, 2000 ; Lim and Dey, 2009 ). Implantation proceeds through three stages: apposition, adhesion/attachment, and penetration ( Carson et al., 2000 ; Schlafke and Enders, 1975 ). During the apposition stage, the trophectoderm comes into close proximity with the luminal epithelium. With uterine lumen closure, firm attachment between the trophectoderm and luminal epithelium is initiated. In mice, the attachment reaction occurs on day 4 of pregnancy at midnight, coinciding with a localized increase of endometrial vascular permeability. Intravenous injection of macromolecular blue dye will clearly mark the sites of blastocyst implantation along the uterine horn due to this vascular permeability ( Lundkvist, 1978a , b , 1979 ; Lundkvist and Ljungkvist, 1977 ). After the attachment reaction, the implanting embryo initiates penetration through the luminal epithelium into the stromal bed ( Carson et al., 2000 ; Schlafke and Enders, 1975 ; Welsh and Enders, 1987 ). Therefore, implantation is a bulk of cell-cell interactions between the blastocyst and the uterus under the primary influence of ovarian steroids ( Figure 3 ). During the preimplantation period, uterine luminal fluid provides a medium for transporting preimplantation embryos into the uterine horn, and absorption of uterine luminal fluid on day 4 of pregnancy facilitates the process of luminal closure and blastocyst apposition within the uterine cavity, establishing intimate contact of the blastocyst and the uterine epithelium ( Figure 1B ). In mice and rats, ovarian estrogen stimulates fluid secretion, while progesterone induces fluid absorption prior to attachment reaction. Interaction between the cystic fibrosis transmembrane conductance regulator (CFTR), a cAMP-activated Cl - channel, and the epithelial Na + channel (ENaC) has been proposed as the major mechanism regulating uterine fluid secretion and reabsorption ( Salleh et al., 2005 ). Previous studies demonstrate that ENaC is primarily localized to the apical membrane of both luminal and glandular epithelia, while CFTR is predominantly expressed in the stromal cells ( Chan et al., 2002 ; Ruan et al., 2012 ; Yang et al., 2004 ). Estrogen induces CFTR expression, but represses ENaC, resulting in fluid accumulation in the uterine lumen, whereas progesterone acts oppositely in regulating these two genes for fluid reabsorption in the uterus ( Nobuzane et al., 2008 ; Zheng et al., 2004 ). Abnormal uterine fluid accumulation and implantation failure have also been observed when CFTR expression is aberrantly upregulated by inflammation in mice ( He et al., 2010 ). The invading embryos can release trypsin, a serine protease known to activate ENaC ( Kleyman et al., 2009 ; Vallet et al., 1997 ). A most recent study demonstrates that the activation of ENaC in the mouse uterus is critical for implantation due to its requirement for PG production and release ( Ruan et al., 2012 ). A related study is conducted on serum and glucocorticoid inducible kinase-1 (SGK1), a key regulator of sodium transport in mammalian epithelia ( Fejes-Toth et al., 2008 ). SGK1 functions through directly activating ENaC and enhancing ENaC expression by inhibiting the ubiquitin ligase, neural precursor cell expressed developmentally down-regulated protein (NEDD) 4–2 ( Lang et al., 2006 ). In mice, Sgk1 mRNA levels transiently decline in the luminal epithelium during the window of endometrial receptivity ( Fisher and Giudice, 2011 ; Salker et al., 2011 ). Intraluminal delivery of an overexpressing Sgk1 vector abolishes normal implantation, with markedly upregulated expression of ENaC α subunit ( Salker et al., 2011 ). In this respect, SGK1 functions as an important enzyme to modulate the appropriate expression and activity of ENaC, which is conducive to uterine fluid absorption prior to implantation. Overall, these findings add a new line of evidence showing that a compensatory expression profile and balanced activity of uterine CFTR and ENaC provide a molecular mechanism, by which maximal fluid absorption and uterine luminal closure can be achieved to facilitate blastocyst apposition. It is thought that adhesion molecules play key roles in blastocyst apposition and attachment during implantation. Previous studies have demonstrated a group of adhesion molecules, such as integrins, the trophinin-bystin-tastin complex, selectins and cadherins, are involved in these processes ( Armant, 2011 ; Kimber and Spanswick, 2000 ; Singh and Aplin, 2009 ). The integrins are versatile and have the ability to bind arginine-glycine-aspartic acid (RGD) sequences in extracellular ligands including fibronectin, osteopontin, vitronectin and others ( Giancotti and Ruoslahti, 1999 ). An attachment mechanism involving the integrins might require a bifunctional bridging ligand to span between the receptors on the embryonic and uterine cell surfaces ( Singh and Aplin, 2009 ). Many integrins have been proposed to have important roles during implantation, including αvβ3, α9β1, αvβ1, α1β1, α3β1, α6β1, αvβ5, and αvβ6 ( Aplin, 1997 ). Among these integrins, the integrin β3 subunit is most broadly studied in both humans and mice. In the human endometrium, the integrin β3 subunit is highly expressed in the luminal and glandular epithelium during the mid-secretory phase, suggesting its potential function in endometrial receptivity ( Lessey et al., 1995 ; Liu et al., 2012 ). Indeed, it has been demonstrated that the aberrant expression of integrin αvβ3 is associated with infertility and recurrent pregnancy loss ( Lessey et al., 1995 ; Liu et al., 2012 ). In mice, αvβ3 is expressed in the uterine luminal epithelium and the blastocyst during implantation, and injection of RGD peptides or neutralizing antibody against the αv or β3 subunit into the mouse uterine cavity reduces the rate of implantation compared to injection of BSA or non-RGD peptides, suggesting its participation in mediating the trophectoderm-luminal epithelium interaction ( Aplin et al., 1996 ; Illera et al., 2000 ; Sutherland et al., 1993 ). There is evidence that integrin signaling modulates trophoblast adhesion to extracellular matrices via activation of phospholipase C-γ to initiate phosphoinositide signaling and intracellular calcium mobilization during blastocyst implantation ( Wang et al., 2002 , 2007b ). “Outside-in” integrin signaling could select appropriate “inside-out” integrin mobilization depending on the extracellular matrix composition ( Armant, 2005 ). Recent evidence shows that integrin β3 expression in the trophectoderm can be repressed by microRNA lethal-7a (let-7a) during blastocyst dormancy with reversal during reactivation for implantation ( Liu et al., 2012 ). Trophinin and tastin form a cell adhesion complex through an intermediary protein bystin that directly binds trophinin and tastin ( Suzuki et al., 1998 ). Previous studies have demonstrated that the trophinin-bystin-tastin complex mediates a unique homophilic adhesion mechanism between trophoblasts and endometrial epithelial cells at their respective apical cell membranes ( Fukuda et al., 1995 ; Suzuki et al., 1998 ). Trophinin, an intrinsic membrane protein, is expressed in trophectoderm cells of embryos and in uterine epithelial cells in mice during the periimplantation period ( Fukuda et al., 1995 ; Suzuki et al., 2000 ). In humans, trophinin expression is restricted to the apical plasma membranes of the endometrial epithelium at the early secretory phase ( Sugihara et al., 2007 ). In a non-human primate model, trophinin is strongly expressed in the trophectoderm of the rhesus macaque blastocyst ( Fukuda et al., 1995 ). Blastocyst-secreted human chorionic gonadotropin (hCG) induces endometrial expression of trophinin, indicative of the spatiotemporal mechanisms that govern embryo-uterine crosstalk for their synchronous development during implantation ( Nakayama et al., 2003 ; Sugihara et al., 2008 ). Bystin is expressed in the luminal and glandular epithelium in the mouse uterus during the periimplantation period and is detected in hatching blastocysts ( Fukuda et al., 2008 ; Fukuda and Nozawa, 1999 ). However, its expression apparently disappears from the blastocyst during implantation, suggesting that its absence is permissive for blastocyst activation ( Aoki et al., 2006 ). Indeed, dissociation of trophinin from bystin by the trophinin-binding peptide, GWRQ, which mimics trophinin homophilic binding, permits activation of ErbB4 to induce trophectoderm activation in human embryo implantation ( Sugihara et al., 2007 ). At the same time, homophilic ligation of trophinin induces tyrosine phosphorylation of PKC-δ and its nuclear translocation in uterine epithelial cells, leading to their apoptosis that is conducive for trophoblast invasion ( Armant, 2011 ). These findings collectively suggest that trophinin, through its dynamic ligation and dissociation from interacting proteins bystin and tastin, potentially mediates the attachment of the blastocyst to uterine epithelial cells at the time of implantation and initiates key physiological changes in both maternal and embryonic cells. L-selectin, a carbohydrate-binding protein, has been proposed as part of a system that mediates the initial adhesion of human blastocysts to the uterine epithelium ( Genbacev et al., 2003 ; Wang et al., 2008 ). L-selectin was previously thought to be expressed only in hematopoietic cells due to its essential role in adhesion between lymphocytes and high endothelial venules ( Berg et al., 1993 ). Intriguingly, human trophoblasts also express functional L-selectin, while its ligand oligosaccharides are detected mainly in pinopodes, the apical cellular protrusions of the endometrial epithelium where blastocyst adhesion initiates ( Genbacev et al., 2003 ; Nejatbakhsh et al., 2012 ). These findings indicate a potential interaction between L-selectin in human blastocysts and oligosaccharide ligands on the endometrial epithelium as an initial step in human implantation. Blocking L-selectin with specific antibodies leads to impaired adhesion of trophoblasts to the endometrial epithelium ( Genbacev et al., 2003 ). Moreover, endometrial expression of L-selectin ligands is significantly different between fertile and infertile women in natural cycles ( Foulk et al., 2007 ; Margarit et al., 2009 ), and impaired expression of L-selectin ligands reduces the chance of successful implantation ( Genbacev et al., 2003 ; Shamonki et al., 2006 ). High expression of L-selectin ligands in the secretory endometrium is likely associated with robust endometrial receptivity for embryo implantation in humans ( Shamonki et al., 2006 ; Wang et al., 2008 ). However, gene knockout mice deficient in fucosyltransferases responsible for the synthesis of fucosylated oligosaccharides ligand of L-selectin are fertile, suggesting that fucosylated carbohydrates, including L-selectin, are dispensable for implantation in mice, although they are expressed in the uterine epithelial cells ( Domino et al., 2001 ). E-cadherin, a Ca 2+ -dependent transmembrane adhesion molecule, mediates intercellular adhesion and dynamic changes to the cytoskeleton through its interaction with cytoplasmic catenins ( Goodrich and Strutt, 2011 ). E-cadherin is a critical factor for blastocyst formation, since embryos lacking the E-cadherin gene fail to establish adhesion junctions in the trophectoderm and die in the periimplantation period ( De Vries et al., 2004 ; Larue et al., 1994 ). On the maternal side, E-cadherin is highly expressed in the luminal epithelium prior to implantation, but is transiently downregulated before blastocyst invasion into the stroma, suggesting that remodeling the adhesion junctions between epithelial cells is a critical event during embryo implantation ( Paria et al., 1999c ; Thie et al., 1996 ; Thie et al., 1995 ; Thie et al., 1998 ). There is evidence that loosening of cell-cell junctions in the mouse uterine epithelium through downregulation of E-cadherin is a prerequisite for blastocyst attachment ( Li et al., 2002 ; Thie et al., 1996 ). Indeed, E-cadherin is persistently expressed in the luminal epithelium of uterine-specific Msx1/Msx2 ablated mice which show implantation failure ( Daikoku et al., 2011 ; Nallasamy et al., 2012 ). Similarly, a recent observation shows that the expression of E-cadherin is remarkably downregulated in endometrial epithelial cells of the mid-secretory endometrium in women with endometriosis ( Matsuzaki et al., 2010 ). Uterine-specific deletion of E-cadherin results in female infertility due to defective implantation and decidualization. The deficient mice lose adhesion junctions and tight junctions in the uterine epithelium, creating a disorganized cellular structure that is incapable of supporting embryo attachment and invasion ( Reardon et al., 2012 ). Collectively, these findings indicate that E-cadherin plays critical roles during embryonic and uterine preparation for implantation. Uterine tissue consists of three major layers: an outer muscle layer, the inner luminal epithelium and the stromal bed in between ( Wang and Dey, 2006 ). Synchronization of estrogen and progesterone directs the uterus into the receptive state, accompanied by morphological and functional changes in the epithelium and the stroma. Increasing attention has been paid to understand how these two hormones execute their differential effects on the two major endometrial cell types, and the molecular basis of stromal-epithelial interactions essential for uterine receptivity ( Cooke et al., 1986 ; Cunha et al., 2004 ; Kurita et al., 2000b ). The synergistic and antagonistic interactions of ovarian progesterone and estrogen on uterine cell proliferation and differentiation are portrayed in Figure 4 . ER is expressed in both epithelial and stromal cells of adult uteri, and it was initially assumed that estrogen exerts its function directly through ER in the corresponding compartments ( Cooke et al., 1998 ). The crucial finding that estrogen stimulates the proliferation of neonatal mouse uterine epithelium, which does not express ER, indicates that estrogen might stimulate uterine epithelial mitogenesis indirectly ( Cooke et al., 1998 ). Employing stroma-epithelium separation/recombination systems ( Cunha, 2008 ) using uteri from adult ERα-deficient mice and neonatal ER intact wild-type mice, a previous study finds that estrogen-induced epithelial proliferation is a paracrine event mediated by stromal ER, not epithelial ER ( Cooke et al., 1997 ). The tissue-specific knockout technique provides an excellent approach to systematically study the uterine responsiveness to estrogen. Selective deletion of ERα in the uterine epithelium (UtEpiaERKO) using Wnt7a-Cre and ER-loxp mouse models demonstrates that stromal ERα is responsible for estrogen-induced epithelial proliferation ( Winuthayanon et al., 2010 ). However, an immediately arising question is how estrogen acting through stromal ERα induces epithelial proliferation. Paracrine actions of polypeptide growth factors, such as insulin-like growth factor (IGF) 1, EGF or TGFα, are believed to be integral components of uterine responses to estrogens. IGF1, a key growth factor induced and activated by estrogen in the uterine stroma, is necessary for estrogen-induced uterine epithelial DNA synthesis through IGF1 receptor signaling in the luminal epithelium ( Chen et al., 2005 ; Kapur et al., 1992 ; Kurita et al., 2005 ; Zhu and Pollard, 2007 ). A previous study has shown that estrogen-stimulated proliferation of uterine epithelial cells is compromised in Igf1 knockout mice, suggesting the role of IGF1 in mediating estrogen action in the endometrium ( Adesanya et al., 1999 ; Sato et al., 2002 ). These studies collectively support a paracrine mechanism of estrogen-mediated epithelial proliferation that solely requires functional ERα in the underlying stroma. After estrogen treatment, PR is dramatically downregulated in the epithelium and increased in the stroma in wild-type and UtEpiaERKO mice, whereas the ER antagonist ICI 182,780 (ICI) inhibits this effect in both genotypes ( Kurita et al., 2001 ; Winuthayanon et al., 2010 ), suggesting that stromal ERα is also required for estrogen-induced down-regulation of uterine epithelial PR. These findings indicate that estrogen acts on stromal ERα to stimulate the proliferation of uterine epithelium through production of paracrine factors, such as IGF1. Although uterine epithelial ERα is dispensable for estrogen-induced epithelial proliferation, it is essential for complete biological and biochemical responses, since selective deletion of uterine epithelial ERα results in compromised uterine weight increase in response to estrogen and epithelial apoptosis after initial proliferation ( Winuthayanon et al., 2010 ). Differentiation of the uterine epithelium, as indicated by secretory products, such as lactoferrin (LF) and Mucin-1, requires functional ERα in both the stroma and the epithelium, and may be a direct effect via ERα plus a paracrine/autocrine effect via synthesis of secreted factors ( Buchanan et al., 1999 ; Kurita et al., 2000a ; Kurita et al., 2000b ). These observations suggest that differentiation of the uterine epithelium requires functional ER in both the epithelium and the stroma. PR-null uteri exhibit a phenotype similar to ovariectomized mice exposed to prolonged estrogen treatment, which is ascribed to an essential modulatory action of PR in the uterus ( Lydon et al., 1995 ). Recombination experiments using uterine tissues from PR-null and wild-type mice demonstrate that stromal PR is required to attenuate estrogen’s proliferative effect on the endometrial epithelium ( Kurita et al., 1998 ). In recent years, numerous genes have been identified that mediate progesterone activity via PR. Immunophilin FK506 binding protein-4 (FKbp52), a co-chaperone required for appropriate uterine PR function ( Daikoku et al., 2005 ), has overlapping expression with PR in uterine stroma ( Tranguch et al., 2005a ). Fkbp52 -/- mice exhibit implantation failure with an exaggerated estrogenic influence in the epithelium ( Tranguch et al., 2005a ; Yang et al., 2006 ). At the histological and cellular level, Fkbp52 -/- uteri display aberrant epithelial proliferation and lower stromal proliferation on day 4 of pregnancy, pointing towards a phenomenon of progesterone resistance ( Tranguch et al., 2005a ; Yang et al., 2006 ). Moreover, ER activity is mostly unaffected and the implantation defect can be rescued by treatment with high dose of progesterone in Fkbp52 null females ( Tranguch et al., 2007 ). Chicken ovalbumin upstream promoter transcription factor II (Coup-TFII, also known as NR2F2), a member of the nuclear receptor superfamily, is highly expressed in the uterine stroma ( Takamoto et al., 2005 ), and its expression is controlled by progesterone-IHH-PTCH1 signaling from the epithelium to the stroma ( Franco et al., 2011b ; Kurihara et al., 2007 ). In addition, uterine conditional knockout of the Coup-TFII gene results in implantation and decidualization failure and enhanced epithelial estrogen receptor activity ( Kurihara et al., 2007 ; Lee et al., 2010 ; Simon et al., 2009 ). These findings suggest that stromal Coup-TFII is an essential PR mediator during implantation and decidualization. The basic helix-loop-helix transcription factor, heart and neural crest derivatives expressed transcript 2 (Hand2), is disclosed by microarray gene profiling analysis of progesterone-responsive transcription in mouse uteri ( Bagchi et al., 2005 ; Li et al., 2011 ). Progesterone induces the expression of Hand2 in the uterine stroma ( Li et al., 2011 ). Selective ablation of Hand2 in uterine cells induces sustained fibroblast growth factor (FGF) expression and its activities in stimulating preimplantation estrogen-induced epithelial proliferation, leading to implantation failure in mice ( Li et al., 2011 ). This finding indicates that Hand2 is a critical regulator of uterine stromal-epithelial communication initiated by steroid signaling that fosters uterine receptivity for implantation. Despite the well-established concept that stromal PR mediates progesterone activity to counter the proliferative response of the epithelium to estrogen, specific roles of epithelial PR in uterine biology have been largely ignored. A most recent study using Wnt7a-Cre/PR loxp mouse models for uterine epithelial PR ablation demonstrates that epithelial PR is essential for uterine epithelial-stromal crosstalk via directly regulating the transcription of progesterone target genes in the epithelium, such as Ihh ( Franco et al., 2011b ). Furthermore, loss of epithelial PR results in a complete pregnancy failure due to impaired decidualization and uncontrolled estrogen-induced epithelial cell proliferation ( Franco et al., 2011b ), highlighting that the epithelium is essential for normal stromal-decidual transformation. This finding clearly demonstrates that epithelial PR is an essential regulator of the stromal-epithelial interaction. In conclusion, progesterone acts through both epithelial and stromal PRs to antagonize the proliferative response of the epithelium to estrogen, while it induces the stromal proliferation. The initial development of the female reproductive tract occurs during embryogenesis soon after gastrulation ( Kobayashi and Behringer, 2003 ; Orvis and Behringer, 2007 ; Spencer et al., 2012 ). In laboratory rodents, domestic animals and humans, the uterus is not fully developed or differentiated at birth; thus, postnatal uterine morphogenesis is a critical event for normal uterine functions in adult life ( Gray et al., 2001a ). Uterine glands are thought to be the principal source of uterine secretions that are required for blastocyst survival and implantation, as well as the establishment and maintenance of pregnancy ( Bazer, 1975 ). Using uterine gland knockout ewe model, it has been demonstrated that uterine gland and its secretion during early pregnancy are essential for conceptus survival ( Gray et al., 2001a ; Gray et al., 2001b ; Gray et al., 2001c ). Molecular and genetic studies have demonstrated that growth factors, cytokine factors and potential histotrophic factors produced by uterine glands are able to regulate the endometrial function. Coordinated development of the endometrial glands from the luminal epithelium, known as uterine adenogenesis, is a predominant event during postnatal uterine morphogenesis. The process of adenogenesis involves three typical steps, budding, penetration and branching. The glandular epithelium derived from the luminal epithelium differentiates and buds, followed by the penetration of the glandular tubes into the uterine stroma, and then the glandular elements keep branching and coiling to expand throughout the endometrial stroma ( Bartol et al., 1999 ; Bartol et al., 1993 ). In mice, epithelial invaginations or gland buds appear at postnatal day (PND) 5, nascent endometrial glands are present at PND 7, and complex glands extending from the luminal epithelium into the endometrial stroma are established by PND 12 ( Branham et al., 1985 ; Brody and Cunha, 1989 ). In contrast, endometrial gland development in humans begins during fetal development within the womb, continues from birth to age 6, and is fully developed by puberty ( Koff, 1933 ). The process of uterine adenogenesis involves intrinsic epithelial-stromal interactions and is precisely orchestrated by multifactorial gene networks ( Baker et al., 1996 ; Miller and Sassoon, 1998 ; Reardon et al., 2012 ). Exposure of mice to native estrogen, estrogen analogs or progesterone during critical developmental periods inhibits endometrial adenogenesis by disrupting the normal expression of morphoregulatory genes ( Cooke et al., 2012 ; Filant et al., 2012 ; Hayashi et al., 2011 ; Miller et al., 1998 ; Stewart et al., 2011 ; Talbi et al., 2006 ). Ablation of these key genes in mice reduces or even eliminates uterine glands ( Table 1 ). Interestingly, abnormalities of uterine glands are often accompanied by defects in the uterine stroma and luminal epithelium, leading to implantation failure. For example, uterine conditional depletion of Dicer, an RNase III endonuclease that is essential for the biogenesis of microRNAs and small interfering RNAs, leads to female sterility and underdeveloped small uteri ( Hawkins et al., 2012 ). The most prominent features of Dicer -deficient uteri are a complete lack of uterine glands and enhanced stromal apoptosis with failure to proliferate in response to progesterone ( Hawkins et al., 2012 ). Females with conditional knockout of prohibitin 2 (also known as repressor of estrogen receptor activity, Rea) in the uterus show implantation and decidualization failure due to altered endometrial gland morphogenesis ( Park et al., 2012b ). In addition, Wnt5a plays a crucial role in the generation of uterine glands and is required for cellular and molecular responses to exogenous estrogen ( Mericskay et al., 2004 ). During postnatal uterine development, Wnt5a participates in a regulatory loop with other uterine patterning genes, including Wnt7a, Hoxa10 and Hoxa11 ( Mericskay et al., 2004 ). Mice with uterine-specific ablation of Wnt4, Wnt7a or E-cadherin show defective implantation associated with a reduction or absence of uterine glands ( Dunlap et al., 2011 ; Franco et al., 2011a ; Hayashi et al., 2011 ). Of the products secreted by uterine glands, LIF is one of the most important molecules. It has been well documented that membrane-associated LIFR and gp130 are required for LIF signaling ( Layton et al., 1994 ; Narazaki et al., 1994 ). In mice, the expression of LIFR is restricted to the luminal epithelium during the preimplantation period and remains highly intense on day 5 when implantation occurs ( Cheng et al., 2001 ; Ni et al., 2002 ; Song and Lim, 2006 ). In contrast, gp130 is expressed in the gland and the stroma on day 4, however, its expression is induced in the luminal epithelium, similar to LIFR, on day 5 ( Bhatt et al., 1991 ; Song and Lim, 2006 ). This overlapping expression pattern of LIFR and gp130 in the luminal epithelium at the attachment site provides robust evidence for the paracrine nature of glandular LIF acting on uterine epithelial cells ( Cheng et al., 2001 ; Stewart et al., 1992 ). Failure of gp130 activation in the luminal epithelium in Lif -/- mice reinforces the notion that LIF-driven glandular-epithelial interaction is essential for normal implantation. Moreover, glandular LIF is essential for the downregulation of Msx1 expression in the uterine luminal epithelium, since the expression of Msx1 persists in the epithelium in Lif null mice during the periimplantation period ( Daikoku et al., 2004 ). Reciprocally, the expression of Lif in the glandular epithelium is also downregulated with loss of uterine Msx1 gene ( Daikoku et al., 2011 ), suggesting that there is a regulatory loop between Lif and Msx1 genes. However, implantation failure resulting from Msx1 deletion cannot be rescued by administration of LIF, suggesting that this mutual regulation is not mediated by the same mechanism ( Daikoku et al., 2011 ). The remarkable positive-negative feedback loop between Msx1 and LIF reflects, to some extent, the complexity of glandular-epithelial interactions during implantation. Since uterine glands are embedded within the stromal bed, it is conceivable that glandular-derived signaling can influence stromal function in a paracrine manner, and vice versa . Forkhead box A2 (Foxa2) is expressed specifically in the uterine glandular epithelium ( Besnard et al., 2004 ). Uterine-specific ablation of this gene in mice results in defective uterine gland formation and dysfunctional uterine stroma unable to support embryo invasion and decidualization ( Bazer, 2010 ; Jeong et al., 2010 ). Notably, LIF expression is significantly decreased in Foxa2 deficient uteri, whereas administration of recombinant LIF can partially restore decidualization in null females ( Bazer, 2010 ; Jeong et al., 2010 ), suggesting that Foxa2 may be required for normal glandular LIF expression. Mice systemically lacking Wnt7a have a stratified luminal epithelium surrounded by a shallow stroma layer devoid of glands, in contrast to the simple columnar structure in the wild-type ( Miller and Sassoon, 1998 ). The conditional uterine ablation of Wnt7a after birth disrupts endometrial gland development, resulting in a failure of blastocyst implantation ( Dunlap et al., 2011 ). Both Foxa2 and LIF are severely reduced in the Wnt7a -/- uterus during the periimplantation period ( Dunlap et al., 2011 ). These findings collectively highlight the necessity of endometrial gland-derived signals for normal stromal function during implantation. The tumor suppressor p53 is crucial for embryonic implantation and decidual development ( Hirota et al., 2010 ; Hu et al., 2007 ). Loss of p53 decreases uterine LIF levels, whereas supplementation of exogenous LIF restores normal implantation in p53 -/- females ( Hu et al., 2007 ). Experiments using human p53 knock-in mice carrying a single nucleotide polymorphism (SNP) at codon 72 (arginine/proline) reveal that mice with the arginine allele exhibit higher uterine LIF levels during implantation than those with the proline allele ( Feng et al., 2011 ), indicating that p53 can regulate the transcription of Lif . However, it is worth noting that p53 is primarily expressed in uterine stromal cells during the periimplantation period ( Yue et al., 2005 ), whereas Lif is initially expressed in the uterine glandular epithelium ( Bhatt et al., 1991 ; Song et al., 2000 ; Stewart et al., 1992 ). These studies hint that signals from the stroma can also influence the physiological function of uterine glands during implantation.

In response to implantation, stromal cells surrounding the implanting embryo undergo extensive proliferation and subsequent differentiation into polyploidy decidual cells ( Lim and Wang, 2010 ; Ramathal et al., 2010 ). The differentiating stromal cells initially form an avascular primary decidual zone (PDZ) encasing the fetus in the afternoon of day 5 ( Paria et al., 1999c ; Wang et al., 2004d ). Stromal cells next to the PDZ continue to proliferate and then differentiate to from a well-vascularized secondary decidual zone (SDZ) ( Figure 1B ). It is believed that a fully developed decidua is important for the provision of nutrition to the developing embryo and to provide a barrier that prevents uncontrolled trophoblast invasion and protects the embryo from the maternal rejection prior to the formation of functional placenta ( Peng et al., 2008 ; Rogers et al., 1983 ; Wang et al., 2004d ; Welsh and Enders, 1987 ). Ovarian progesterone signaling through PRA is essential for postimplantation uterine development and decidualization ( Conneely et al., 2002 ). However, the potential role of PRB during decidualization cannot be ruled out since mice lacking both PRB and p21, a cyclin dependent kinase (CDK) inhibitor, are infertile with decidualization defects ( Das, 2009 ). In addition to FKBP52, transcriptional coregulators are also crucial in mediating PR activity, such as the steroid receptor coactivator (SRC)-1, SRC-2 and SRC-3 ( McKenna and O’Malley, 2002 ; Spencer et al., 1997 ). These cofactors have certain overlapping and distinct physiological functions in the female reproductive system. For example, mice with a null mutation of either SRC-1 or SRC-2 show progesterone resistance and comprised decidualization, whereas mice lacking both SRC-1 and SRC-2 are infertile with a complete blockage of decidualization ( Jeong et al., 2007 ; Mukherjee et al., 2006a ; Mukherjee et al., 2006b ; Xu and Li, 2003 ; Xu et al., 1998 ). In contrast, SRC-3 knockout mice display delay puberty, but have normal decidualization ( Xu et al., 2000b ). Interestingly, SRC-2 also acts as a coactivator of PPARδ, a nuclear hormone receptor activated during stromal differentiation by COX-2-derived prostacyclin during implantation and decidualization ( Lim et al., 1999a ; Lim et al., 2004 ). Since progesterone alone is adequate to maintain the decidual response in experimentally induced decidualization after ovariectomy ( Paria et al., 1999b ), it was thought that estrogen is dispensable in this process. Conversely, decidualization is compromised by administration of ICI after embryo attachment, suggesting a contribution of an intrauterine source of estrogen in the decidua ( Curtis et al., 1999 ). Moreover, later studies show that blocking P450 aromatase, a key enzyme that converts androgen to estrogen and is expressed in decidual stromal cells ( Simpson et al., 1994 ), inhibits the process of decidualization, suggesting that de novo synthesized estrogen in the uterus is essential for normal decidualization ( Das et al., 2009 ). As a downstream target of estrogen, Fos-related antigen 1 (FRA-1), a member of the Fos family of transcription factors, has recently been shown to be a critical estrogen mediator during decidualization ( Das et al., 2012 ). FRA-1 is expressed in the differentiating uterine stroma surrounding the implanted embryo and this expression is significantly repressed in response to letrozole. Loss of FRA-1 inhibits stromal cell differentiation and migration ( Das et al., 2012 ). By employing rodent models with experimentally ablated uterine epithelium, evidence has been produced that the uterine luminal epithelium transmits signals required for the induction of decidualization ( Lejeune et al., 1981 ). After blastocyst attachment, the luminal epithelial cells surrounding the invading blastocyst undergo apoptosis ( Parr et al., 1987 ; Schlafke et al., 1985 ; Welsh and Enders, 1991 ) to facilitate embryo invasion and access to maternal blood supply ( Pampfer and Donnay, 1999 ; Parr et al., 1987 ). The apoptotic degeneration of the uterine epithelium involves embryo-derived signaling molecules, such as transforming growth factor β ( Kamijo et al., 1998 ; Mahesh et al., 1996 ) and trophinin ( Tamura et al., 2011 ), and is mediated by caspase 3 ( Joswig et al., 2003 ; Zhang and Paria, 2006 ). Recent studies have demonstrated that KLF5, a zinc finger-containing transcription factor is primarily expressed in the luminal epithelium throughout the periimplantation stage ( Ema et al., 2008 ; Sun et al., 2012 ). Conditional ablation of Klf5 leads to implantation failure, suggesting that KLF5 contributes to the establishment of uterine receptivity. In addition, Klf5 -null mice show decidualization defects, primarily due to a failure of the trophoblast cells to penetrate the luminal epithelium ( Sun et al., 2012 ). Conversely, premature death of endometrial epithelium from aberrantly upregulated expression of CFTR in response to high levels of estrogen contributes to impaired embryo implantation ( Yang et al., 2011 ). These findings highlight the necessity of appropriately differentiated luminal epithelium for inducing the decidual response. With the advance of uterine specific gene depletion models, a number of key regulators have been identified involving in epithelial-stromal interactions that initiate decidualization ( Lim and Wang, 2010 ; Ramathal et al., 2010 ; Wetendorf and DeMayo, 2012 ; Yang et al., 2011 ) ( Figure 5 ). Ihh is expressed in the uterine epithelium under the control of progesterone ( Lee et al., 2006 ; Matsumoto et al., 2002 ). In addition to its essential role in uterine receptivity, Ihh -null mice also exhibit decidualization failure in response to an artificial stimulus ( Lee et al., 2006 ). Since both the IHH receptor Ptch1 and the downstream effectors, Gli1–3 and COUP-TFII, are expressed in the uterine stroma, it is presumed that IHH signals from the luminal epithelium to the stroma through Ptch1 and activates downstream transcription factors. Indeed, ablation of COUP-TFII also results in severe decidual defects, suggesting that this cascade is required for decidualization ( Kurihara et al., 2007 ; Lee et al., 2006 ). Interestingly, the decidual defects in these mice can be partially rescued by intraluminal supplementation of recombinant BMP2, suggesting that BMP2 also participates in IHH signaling and operates as a downstream effector of COUP-TFII in decidualization ( Kurihara et al., 2007 ). Collectively, these studies indicate that the progesterone-induced IHH/Ptch1/COUP-TFII/BMP2 pathway mediates the epithelial-stromal crosstalk that is critical for implantation and decidualization. Polyploidization is a hallmark of decidualization that utilizes a specialized cell cycle progression ( Das, 2009 ; Dey et al., 2004 ). It is speculated that polyploidy limits the life span of decidual cells, but ensures that these cells meet the demand for postimplantation embryonic growth ( Ansell et al., 1974 ; Davoli and de Lange, 2011 ; Hemberger, 2008 ; Sachs and Shelesnyak, 1955 ). Cyclin D3 is a key cell cycle regulator in stromal cell proliferation, differentiation and polyploidization ( Das, 2009 ; Das et al., 1999 ). Its deficiency in mice leads to defective decidualization ( Das, 2009 ; Das et al., 1999 ). Moreover, previous study has identified several other regulators crucial for decidualization executing their function through cyclin D3. For instance, female deficiency of Hoxa10 , which is highly expressed in developing decidua, exhibits impaired decidualization with aberrant regulation of cyclin D3 and loss of region-specific expression of CDK4 and CDK6 in the decidua bed ( Rahman et al., 2006 ). Female deficiency of basic transcription element-binding protein 1 (BTEB1), a potential upstream regulator of Hoxa10, also exhibits decidualization defects with dysregulation of both Hoxa10 and cyclin D3 expression in the uterus ( Pabona et al., 2012 ; Simmen et al., 2004 ). Interleukine-11 receptor α (IL-11Rα) is the α-chain of the receptor for IL-11. A null mutant of this gene in mice results in decidual degeneration with derailed endoreplication due to a reduced cyclin D3 expression ( Bilinski et al., 1998 ; Li et al., 2008 ; Menkhorst et al., 2009 ; Robb et al., 1998 ). In addition, p21 and survivin are downstream targets of IL-11-stimulated decidualization ( Li et al., 2008 ). Observation of decidual defects in mice lacking both p21 and PRB indicates a critical role of p21 in decidualization ( Li et al., 2008 ). A recent study has demonstrated that the death effector domain-containing protein (DEDD), which can stabilize cyclin D3, is essential for normal uterine decidualization. Its deficiency leads to impaired decidual development accompanied by attenuated polyploidy ( Arai et al., 2007 ; Mori et al., 2011 ). HB-EGF also facilitates stromal cell polyploidy by upregulating cyclin D3 expression ( Das, 2009 ; Tan et al., 2004 ). All these findings point towards the central role of cyclin D3 during decidual cells proliferation and polyploidization. In addition, many cell-cycle associated genes, such as hepatoma upregulated protein (Hurp) and CCAAT/enhancer binding protein β (C/EBPβ), have also been proved to be essential for decidual development ( Tsou et al., 2003 ). Hurp is highly expressed in G2/M phase of the cell cycle and participates in regulating spindle formation and chromosome congression during mitosis ( Bassal et al., 2001 ). Female mice lacking Hurp are completely infertile due to a failed decidual reaction caused by a defect in stromal proliferation that leads to implantation failure ( Tsai et al., 2008 ). C/EBPβ is rapidly induced in the stromal cells surrounding the implanted blastocyst and the expression is highly increased during decidualization. Ablation of C/EBPβ gene in female mice results in infertility with a complete lack of decidual formation ( Bagchi et al., 2006 ; Ramathal et al., 2011 ; Wang et al., 2010a ). Further studies show that C/EBPβ functions during decidualization through modulating the expression of multiple key cell cycle regulatory factors that control G2 to M transition of the proliferating stromal cells ( Bagchi et al., 2006 ; Ramathal et al., 2011 ; Wang et al., 2010a ) ( Figure 5 ). Decidua is a privileged site for coordinating the process of immune tolerance to protect the genetically ‘foreign’ semi-allogeneic embryo from attack by the maternal immune system. Indeed, the avascular PDZ, composed of several layers of semi-epithelialized stromal cells, acts as a physical barrier to restrict the access of immunologically competent maternal cells and/or their secretions ( Tung et al., 1986 ; Wang et al., 2004d ). A large number of molecules and immune cell types participate in immune surveillance and tolerance during pregnancy ( Dosiou and Giudice, 2005 ; Moffett and Loke, 2006 ; Taglauer et al., 2010 ; Trowsdale and Betz, 2006 ; Yoshinaga, 2012 ) ( Figure 5 ). The maternal circulating tryptophan concentration falls with the progression of pregnancy and the establishment of maternal tolerance, and this tryptophan catabolism can suppress the T cell activity and defend the conceptus against maternal rejection ( Fuchs et al., 1996 ; Kamimura et al., 1991 ; Munn et al., 1996 ; Munn et al., 1998 ). The first clear example of allospecific abortion during early pregnancy comes from the study on the effect of 1-methyl-tryptophan, a pharmacologic agent that inhibits the activity of indoleamine 2,3-dioxygenase (IDO) enzyme, which catabolizes tryptophan, on murine pregnancy ( Cady and Sono, 1991 ; Munn et al., 1998 ). A systemic administration of 1-methyl-tryptophan that causes continuously high tryptophan concentration induces fetal deterioration by embryonic day 9.5 in female mice mated with allogeneic males but not with syngeneic males ( Munn et al., 1998 ). IDO is expressed in dendritic cells, macrophages and trophoblasts in mice and humans ( Munn et al., 1998 ; Saito et al., 2007 ), and its activity is also detectable in the decidua ( Jeddi-Tehrani et al., 2009 ; Kudo et al., 2004 ). A high tryptophan-degrading activity is present in the early conceptus on days 7 and 8 of pregnancy in mice, whereas IDO is not expressed during this early gestational stage, but appeared 2–3 days later concomitant with the formation of the placenta ( Suzuki et al., 2001 ). Tryptophan 2,3-dioxygenase (TDO), another key catabolizing enzyme of tryptophan, contributes to this early tryptophan-degrading activity ( Suzuki et al., 2001 ). Expression of TDO is induced in the decidualizing stroma surrounding the implanted embryo. With the progression of pregnancy, its expression is further upregulated and shifts to the mesometrial decidua, a site where future placenta will form ( Suzuki et al., 2001 ; Tatsumi et al., 2000 ).These findings demonstrate that TDO and IDO present in the endometrium and trophoblasts concomitant with decidualization is necessary to prevent immunological rejection of fetal allografts during pregnancy recognition and maintenance. Moreover, 1-methyl-tryptophan- induced abortion occurs when the T and B cell-deficient ( Rag-I -/- ) mother carries CD8 + T cells specific for a fetal major histocompatibility complex (MHC) antigen inherited from the father, whereas does not occur when the mother carries syngeneic fetuses, suggesting that this fetal allograft rejection triggered by 1-methyl-tryptophan is T cell-dependent and antibody-independent activation of complement deposition ( Mellor et al., 2001 ). Interestingly, IDO-deficient females produce litters of normal sizes at normal rates, and resistance to 1-methyl-tryptophan injection ( Baban et al., 2004 ), implying that some other molecules, such as uterine TDO, elicit the compensatory function to protect the allogeneic fetuses. Alternatively, this phenomenon can be argued that abortion induced by IDO-catabolized metabolic products from 1-methyl-tryptophan is allospecific and T cell-dependent. Other molecules associated with immune tolerance have been found in succession, such as programmed death ligand 1 (PDL1). PDL1 is an inhibitory T cell costimulatory molecule, which has been proved to be pivotal in regulating immune responses in many systems ( Khoury and Sayegh, 2004 ; Rothstein and Sayegh, 2003 ). Interestingly, its expression in the uterus is restricted to the decidua basalis ( Guleria et al., 2005 ). This specific expression pattern suggests a role of PDL1 in regulating the maternal alloimmune response. Blockage of PDL1 signaling with specific monoclonal antibody leads to a significant increase in the rejection rate of allogeneic concepti in B cell-deficient but not in Rag-I -/- mothers, and therefore this fetal rejection was T cell- but not B cell-dependent ( Guleria et al., 2005 ; Guleria and Sayegh, 2007 ). Furthermore, ablation of PDL1 in female mice results in severe reduction in allogeneic fetal survival rates ( Guleria et al., 2005 ; Guleria and Sayegh, 2007 ). These observations together indicate that the PDL1 is crucial in fetomaternal immune tolerance. Galectins (Gal) are a family of β-galactoside-binding lectins which execute their immune functions through orchestrating the proliferation and survival of effector T cells, and among them, Gal-1 can directly induce the death of T cells, which is favorable to fetal survival in the maternal uterus ( Blaser et al., 1998 ; Chung et al., 2000 ; Rabinovich et al., 2007 ; Stillman et al., 2006 ). In mouse uteri, the expression of Gal-1 is regulated by ovarian steroids and increased in the decidua during early pregnancy ( Hirota et al., 2012 ). Gal-1 -deficient mice show higher rates of fetal loss just when the female mice are mated with the allogeneic males ( Blois et al., 2007 ). A functional crosstalk between the immune and endocrine systems exists in the regulation of fetomaternal tolerance ( Arck et al., 2007 ). For example, Gal-1 treatment prevents the stress-induced decrease of circulating progesterone ( Blois et al., 2007 ). Reciprocally, treatment with progesterone increases Gal-1 expression in the myometrium and decidua of stress-challenged pregnancies ( Blois et al., 2007 ). Previous studies have shown that exogenous progesterone can rescue implantation failure in Fkbp52 -/- mice, but the embryo resorption rate remains high at midgestation ( Tranguch et al., 2005a ; Tranguch et al., 2007 ; Yang et al., 2006 ). This is consistent with recent findings that Fkbp52 deficiency results in significantly down-regulated Gal-1 expression, whereas concomitant supplementation of recombinant Gal-1 can suppress the high incidence of fetal resorptions in Fkbp52 -/- females ( Hirota et al., 2012 ). These results provide robust evidence for the crosstalk between progesterone-FKBP52-PR signaling and Gal-1 in immune surveillance during midgestation pregnancy. Complement activation is a component of innate T-cell driven immunity and its accurate regulation is necessary for protecting tissues from inflammation ( Mellor et al., 2001 ; Molina, 2005 ). Complement component receptor 1-like protein (Crry), a murine negative regulator of complement activation, is highly expressed in the trophoblast cells as early as embryonic day 7.5, as well as in the decidual tissues during early pregnancy ( Hagmann, 2000 ; Xu et al., 2000a ). Systemic deletion of Crry leads to embryonic lethality due to abnormal complement deposition in the placenta ( Mellor et al., 2001 ; Molina, 2005 ; Xu et al., 2000a ). However, when this mutation is backcrossed into complement component 3 (C3)-deficient background, the knockout fetus can survive until adulthood, indicating that the Crry -/- embryonic lethality is resulted from the activation of complement system ( Xu et al., 2000a ). It is further proved that the Crry -/- embryonic lethality is mediated exclusively by the activation of maternal C3, independent of another component C5 ( Mao et al., 2003 ). Moreover, C3 activation has been identified as the main cause of pathogenesis of antiphospholipid syndrome with recurrent fetal loss in humans ( Holers et al., 2002 ). Stromal differentiation and angiogenesis during decidualization are crucial for pregnancy establishment and maintenance. Decidual leukocytes are thought to play a role not only in creating maternal immune tolerance, but also in decidual development and remodeling during early pregnancy. Leukocytes comprise a large part of cell population in decidual bed during early pregnancy both in humans and mice ( Croy et al., 2012 ; Lash et al., 2010a ). Among these leukocytes, the major player involved in maternal immune tolerance is the uterine natural killer (uNK) cells, which become activated and dramatically increased during decidualization, accounting for about 65–70% of the decidual immune cells ( Croy et al., 2002 ; Dosiou and Giudice, 2005 ; Lash et al., 2010b ). Unlike T and B lymphocytes, uNK cells are devoid of somatically rearranged antigen-sensing receptors ( Natarajan et al., 2002 ), but express CD56 brightly and lack CD16 and L-selectin, which distinguish them from peripheral NK cells ( Searle et al., 1999 ). uNK cells proliferate rapidly between days 5–7 of pregnancy in mice within the uterus, where they execute the differentiation program under the control of local microenvironment ( Chantakru et al., 2002 ; Croy et al., 2006 ). Although murine trophoblast cells express target molecules recognized by the uNK cells ( Chatterjee-Hasrouni et al., 1984 ), they remain unharmed in the presence of the decidual uNK cells, which is also blocked by prostaglandin secreted from the decidual cells ( Parhar et al., 1989 ). The physiological role of uNK cells is to provide growth support for stromal cells to differentiate into decidual cells by modifying spiral arteries ( Charalambous et al., 2012 ; Greenwood et al., 2000 ; Hanna et al., 2006 ; Herington and Bany, 2007b ), which are critical for sufficient fetus nutrition supply ( Adamson et al., 2002 ). Ablation of uNK cells in mice often induces severe abnormalities in the uterus, such as absence of metrial gland, hypocellularity and edema of decidual bed and blockade of normal modification of spiral arteries ( Ashkar et al., 2000 ; Guimond et al., 1998 ). This defect is uNK-specific as bone marrow transplants from severe combined immunodeficient (SCID) mice, which are T and B cell-deficient, can rescue this phenotype ( Guimond et al., 1998 ). In humans, extravillous cytotrophoblast cells express a unique combination of MHC class I molecules, namely human leukocyte antigen (HLA)-C, -E and -G, and among these only HLA-C is polymorphic ( Moffett-King, 2002 ). Another crucial phenotype of uNK cells is that they express highly diverse killer immunoglobulin-like receptors (KIRs) that recognize polymorphic HLA-C. Different KIR and HLA-C haplotype combinations can limit or foster uNK activation and uterine vasculature remodeling, and notably, are associated with the etiology of pre-eclampsia ( Hiby et al., 2010 ; Redman and Sargent, 2005 ). Antigen-presenting cells (APCs) are the second abundant leukocyte population in the decidua, displaying foreign antigen complexes with MHC on their surfaces, and these complexes bind to T-cell receptors (TCRs) to activate T cells ( Cella et al., 1997 ). Among APCs, CD11c hi dendritic cells (DCs) expressing both MHC class I and II molecules are the most potent inducers of immune responses. DCs have been confirmed to exist in the nonpregnant and pregnanant endometrium in both humans and rodents ( Blois et al., 2004 ; Gardner and Moffett, 2003 ; Zarnani et al., 2007 ). In mice, DCs are located subadjacent to the luminal epithelium at early pregnancy, and randomly distributed in the stroma and around the epithelium afterwards ( Zarnani et al., 2007 ). Inducible ablation of DCs can be achieved by the administration of diphtheria toxin (DT) in transgenic mice carrying diphtheria toxin receptor (DTR) under the control of CD11c promoter, and therefore the CD11c-expressing DC cells are susceptible to acute ablation ( Jung et al., 2002 ; Saito et al., 2001 ). By using this approach, several groups demonstrate that depletion of DCs impairs the proliferation and differentiation of endometrial stromal cells during decidulization ( Krey et al., 2008 ; Plaks et al., 2008 ; Pollard, 2008 ). Moreover, abalation of DCs also affects the accumulation of mature uNK cells ( Krey et al., 2008 ). An impaired trafficking of CD11c + CD11b + DC cells to the pregnant uterus are accompanied with aberrant uNK differentiation in lower numbers and with smaller cell size ( Behrends et al., 2008 ; Karsten et al., 2009 ), suggesting that DC-derived signals may directly mediate uNK homing and maturation in the pregnant endometrium, as reviewed ( Blois et al., 2011 ). A milestone for understanding the fetomaternal immune tolerance is a unique Act-mOVA transgenic mouse model, which express the transmembrane form of ovalbumin (OVA) under the control of the ubiquitous β-actin promoter. Unexpectly, the OVA transgene is expressed in the labyrinthine trophoblasts and particularly by invasive trophoblasts at the fetomaternal interface ( Ehst et al., 2003 ; Erlebacher et al., 2007 ). As a result, the nontransgenic females bearing the transgenic conceptus will encounter the antigen OVA, which serves as a surrogate fetal or placental alloantigen and allows for visualizing antigen prsentation at mouse fetomaternal interface ( Erlebacher et al., 2007 ). Employing this model, Erlebacher et al. demonstrate that T cells have a specific and robust proliferative responses in female mice bearing OVA-expressing conceptus ( Erlebacher et al., 2007 ). However, the OVA antigen presentation is mediated exclusively by APCs of maternal rather than fetal origin, and this is important because allorecognition relying on maternal APCs is minor and avoids a major threat to fetal survival by T cells recognizing fetal MHC complexes ( Erlebacher et al., 2007 ; Moldenhauer et al., 2009 ). Moreover, CD8 + cytoxic T cells that indirectly recognize fetal antigen can undergo clonal deletion and cannot prime for their effector function ( Erlebacher et al., 2007 ). More strikingly, the decidual DCs become trapped and immobile within the pregnant uterus, and as a result, these resident DCs cannot reach the lymphatic vessels, which solely exist in the myometrium ( Chakraborty and Pulendran, 2009 ; Collins et al., 2009 ). Therefore, maternal T cells appear to be ignorant of fetal antigens and immnue surveilance at the fetomaternal interface, as comprehensively reviewed by Erlebacher recently ( Erlebacher, 2010 ). Regulatory T cells (Treg cells), also known as suppressor T cells, are a population of CD4 + CD25 + T cells, which contribute to maintaining in vivo tolerance by suppressing immune response in an antigen-nonspecific manner ( Sakaguchi et al., 1995 ; Shevach, 2011 ). Such immune activities of Treg cells suggest their involvement in impeding the maternal allogeneic response against the fetal antigen. Recent studies suggest that circulating and uterine CD4 + CD25 + Treg cells increase during pregnancy in both humans and mice ( Aluvihare et al., 2004 ; Heikkinen et al., 2004 ; Robertson et al., 2009 ; Sasaki et al., 2004 ; Zenclussen, 2005 ; Zhu et al., 2005 ). Therefore, it is not surprising that absence of Treg cells will lead to pregnancy loss due to uncontrolled immune rejection of the fetus. For example, in Treg-deficient BALB/c nude (nu/nu) females, transferring Treg-depleted polyclonal T cells leads to fetal resorption in allogeneic pregnancy ( Aluvihare et al., 2004 ). Consistently, Treg depletion with anti-CD25 monoclonal antibodies during implantation phase or early pregnancy also results in implantation failure or rapid embryo resorption in allogeneic pregnancy ( Darrasse-Jeze et al., 2006 ; Shima et al., 2010 ), whereas Treg cells are dispensible for the maintenance of the late stage of maternal tolerance to the allogeneic fetus ( Shima et al., 2010 ). Notably, seminal fluid at mating also contributes to maternal immune tolerance to paternal alloantigens via mediating the expansion of Treg cell pool in the uterine draining lymph nodes at the onset of pregnancy ( Robertson et al., 2009 ). In addition to CD4 and CD25, endogenous Treg cells also specifically express the forkhead box P3 (Foxp3), the key control gene in their development ( Fontenot et al., 2003 ; Hori et al., 2003 ), and Foxp3 dysfunction causes autoimmune disease in humans and mice ( Sakaguchi, 2005 ). Ablation of Foxp3-expressing Treg cells with diphtheria toxin also fractures maternal tolerance to fetal antigen and triggers fetal resorption with robust expansion of fetal antigen-specific CD4 + and CD8 + T cells ( Rowe et al., 2011 ). In humans, a decrease of Treg cell population and their expression of transcript factor Foxp3 has been reported in paitents with miscarriage or unexplained infertility ( Jasper et al., 2006 ; Sasaki et al., 2004 ; Yang et al., 2008 ). Remarkably, conserved noncoding DNA sequence 1 (CNS1) present only in placental mammals, has recently been characterized as a Foxp3 enhancer ( Zheng et al., 2010 ), and CNS1 -deficient females exhibit increased fetal resorption in allogeneic pregnancy, resulting from failed accumulation of paternal alloantigen-specific Treg cells in the decidua ( Samstein et al., 2012 ). During the pregnancy, sustained expansion of Treg cells contributes to the maternal immune tolerance, thus it seems that the ability of host defense against infection will reduce ( Rowe et al., 2011 ; Rowe et al., 2012a ). A recent study highlights the specificity of Treg cells during pregnancy, which uncovers the myth to dissociate their beneficial and detrimental impacts. Treg cells have a memory after the first pregnancy and these fetal-specific Treg cells can maintain tolerance to pre-existing fetal antigen more rapidly and effectly, which could be resistant to partial maternal Foxp3 cell ablation ( Rowe et al., 2012b ). The T helper type 1 (Th1)-Th2 theory proposed in 1986 ( Mosmann et al., 1986 ) defines that Th1 cells-derived cytokines such as IFN-γ predominantly promote cell-mediated immunity, whereas cytokines produced by Th2 cells such as IL-4 induce humoral immunity and the production of immunoglobulin antibodies. It has been once considered that successful pregnancy is dependent upon the Th1/Th2 balance, in that Th2 cytokines favor the fetal growth whereas Th1 cytokines hamper it ( Piccinni and Romagnani, 1996 ; Saito, 2000 ; Zenclussen, 2005 ; Zenclussen et al., 2005 ). T cells from the decidua of women with miscarriage can produce predominant Th1-type cytokines with decreased Th2-type cytokines ( Piccinni et al., 1998 ; Piccinni and Romagnani, 1996 ; Raghupathy et al., 1999 ), and this is also the case in mice with impaired gestation ( Zenclussen, 2005 ). Moreover, previous studies show that Th1/Th2 cytokines can regulate uterine expression of FAAH and LIF and trophoblast release of hCG ( Karasu et al., 2011 ; Saito, 2000 ), all of which are known to play roles during implantation. However, the Th1-Th2 paradigm is questioned since IFN-Y that is detrimental to human pregnancy facilitates uterine vascular modification, decidualization and uNK cells differentiation in mice ( Ashkar et al., 2000 ). Mice lacking Th2-type cytokines IL-10 and IL-4 show apparently normal pregnancy ( Svensson et al., 2001 ), further shaking the Th1-Th2 theory. In fact, this traditional Th1-Th2 paradigm has evolved into a more complicated theory. Recent evidence shows that naive CD4 + T helper cells differentiate into four effector subsets, Th1, Th2, Th17 and Treg cells, which secret distinct combinations of cytokines ( Tato and O’Shea, 2006 ). It is becoming extensively recognized that cytokines are of multiple cellular originations, rather than solely secreted from one subset of leukocyte ( Coffman, 2006 ). This view helps to resolve the longstanding paradox of Th1-Th2 theory. Adaptive immune defense tailored to different pathogens is orchestrated by Th1, Th2 and Th17, whereas Treg cells perform an immunosuppression role, as discussed above. The development of these subsets is under the control of distinct lineage specifying genes. For example, STATs, T-box transcription factor (T-bet) and GATA binding protein 3 (GATA-3) are crucial for the differentiation along the Th1 and Th2 pathways, while retinoic acid receptor (RAR)-related orphan nuclear receptor RORYt is the master regulator of Th17 cells, as reviewed excellently ( Lehar and Bevan, 2004 ; Reiner, 2007 ; Tato and O’Shea, 2006 ). Notably, a very recent study shows the differentiation of Th2 lineage is controlled by an epigenetic transcriptional silencing pathway ( Allan et al., 2012 ). The Th1/Th2/Th17 and Treg paradigm in implantation has also drawn extensive attention of reproductive immunologists ( Lee et al., 2012b ; Saito et al., 2010 ). Increasing evidence shows an elevated prevalence of IL-17-producing cells and/or Th17 immune response in patients with pregnancy complications, miscarriage ( Nakashima et al., 2012 ; Wang et al., 2010b ) and preeclampsia ( Darmochwal-Kolarz et al., 2012 ; Darmochwal-Kolarz et al., 2002 ; Santner-Nanan et al., 2009 ; Toldi et al., 2011 ). Similar as Th1 cells, Th17 cells tend to reject the conceptus alloantigens and are thus harmful to pregnancy maintenance in humans ( Liu et al., 2011 ). Interestingly, progesterone suppresses the development of Th17 cells via inhibiting STAT3 activation in response to IL-6 ( Lee et al., 2011b ). However, a recent study fails to observe a predominant Th1 and Th17 response in women with repeated IVF failure ( Persson et al., 2012 ), suggesting a diversified etiology of gestation diseases. Overall, the hitherto useful Th1/Th2/Th17 and Treg paradigm can be expanded into a well-orchestrated cytokine network conducive for maternal immune tolerance during pregnancy recognition and adaption. Effector T cells are the main cell type causing immune rejection owing to their cytolytic and cytokine-producing effector function. At the fetomaternal interface, as mentioned, decidua-entraped DC cells contribute to the prevention of fetal rejection during early pregnancy via minimizing the activation of naive T cells ( Collins et al., 2009 ; Erlebacher et al., 2007 ). Moreover, the fetus can aviod the T cell attackment even when antigen-specific cytotoxic T lymphocyte (CTL) activity is experimentally induced in late gestation ( Collins et al., 2009 ), pointing towards the existence of a fail-safe mechanism protecting the conceptus from activated CTLs. In support of this, a recent study demonstrates that effector T cells fail to infiltrate into the decidua in response to fetal challenge, and instead are trapped within the myometeial layer of the pregnant uteri ( Nancy et al., 2012 ; Reinhardt et al., 2003 ). This impaired accumulation of cytolytic T cells in the decidua is in part attributable to the epigenetic silencing of key T cell-attracting chemokine genes, and as a result the process of decidualization reduces the capacity of the decidual cells to produce T cell chemoattractants ( Nancy et al., 2012 ; Reinhardt et al., 2003 ). For example, the expression of chemoattractants for T cell infiltration, such as CCL5 and CXCL9, are synergistically induced in the myometrium, whereas their expression remains at low levels in the dedidua, via a repressive promoter regulatory status presented by the histone H3 trimethyl lysine 27 (H3K27me3) mark ( Nancy et al., 2012 ; Reinhardt et al., 2003 ). In humans, T cells have also been reported to be relatively rare in the decidua, and chemoattractant CXCL10 is expressed only locally in decidual tissues, in association with periglandular leukocyte aggregations ( Bulmer et al., 2010 ; Marlin et al., 2011 ; Red-Horse et al., 2001 ; Volchek et al., 2010 ).

Most of what is understood about the molecular signaling pathways that govern embryo-uterine interactions has been obtained using animal models because human experimentation is limited by ethical considerations and restrictions on research with human embryos ( Berlanga et al., 2011 ; Lim and Wang, 2010 ). Although humans and rodents are evolutionarily different, they share many physiological similarities with respect to embryo implantation. For example, blastocyst hatching or zona pellucida dissolution occurs in both species to ensure proper contact and orientation of the blastocyst within the uterine lumen ( Chen et al., 2012 ). After blastocyst apposition and spacing, blastocyst attachment initiates a precisely regulated embryo-uterine dialogue that occurs within a defined window of implantation in both species ( Bazer et al., 2009 ; Singh and Aplin, 2009 ; Wang and Dey, 2006 ). Stroma cells underlying the luminal epithelium undergo extensive decidualization upon trophoblast penetration to protect and support the development of embryos in both humans and mice, although stromal-decidual transformation begins independently of an embryo in women during the mid-secretory phase of the menstrual cycle ( Bazer et al., 2009 ). These similarities strengthen the value of the tremendous advances in implantation research achieved using mouse models. Many signaling molecules found to be crucial for embryo implantation in mice are of high clinical significance for human implantation and pregnancy outcome ( Table 2 ). For example, LIF, a crucial cytokine in mouse implantation, appears essential in humans, since LIF gene mutations are often associated with unexplained infertility and repeated implantation failures ( Aghajanova, 2004 ; Aghajanova et al., 2009 ; Steck et al., 2004 ). Moreover, women with high endometrial LIF immunoreactivity during the window of implantation are more likely to become pregnant than those with weaker expression in IVF-ET program ( Serafini et al., 2009 ). Anandamide at high levels is detrimental to blastocyst differentiation and uterine receptivity in mice ( Saito, 2000 ; Sun and Dey, 2009 ; Wang et al., 2006a ). Clinical evidence suggests that high peripheral levels of anandamide are correlated with recurrent pregnancy losses in women ( Bari et al., 2011 ; Habayeb et al., 2008 ; Maccarrone, 2009 ; Maccarrone et al., 2002 ; Maccarrone et al., 2000 ). While the lipid signaling cascade of LPA3-cPLA2α-COX-2-PG is essential for normal implantation and decidualization in mice ( Lim et al., 1999a ; Lim et al., 1997 ; Song et al., 2002 ; Wang et al., 2004b ; Wang et al., 2007a ; Ye et al., 2005 ), the physiological significance of PG and LPA signaling in human embryo implantation is supported by observations that women with repeated implantation failure have reduced levels of PG and LPA ( Achache et al., 2010 ; Marions and Danielsson, 1999 ). HB-EGF activation of ErbB1/4 receptors involves bidirectional interactions between the blastocyst and the uterus prior to implantation in mice ( Das et al., 1994 ; Hamatani et al., 2004b ; Paria et al., 1999a ; Paria et al., 2001a ; Raab et al., 1996 ; Wang et al., 2000 ; Xie et al., 2007 ), and a similar machinery functions during human implantation and placentation ( Birdsall et al., 1996 ; Chobotova et al., 2005 ; Chobotova et al., 2002a ; Chobotova et al., 2002b ; Gonzalez et al., 2011 ; Jessmon et al., 2009 ; Leach et al., 1999 ; Lessey et al., 2002 ; Yoo et al., 1997 ). Moreover, aberrant expression of HB-EGF in human trophoblasts is associated with preeclampsia ( Aghajanova et al., 2008a ; Leach et al., 2002 ). In addition, homeobox transcription factors and developmental morphogenesis genes essential for uterine function and implantation in mice exhibit dynamic expression patterns in human endometrium and potential association with certain pregnancy complications ( Bombail et al., 2010 ; Cheng et al., 2008 ; Kodama et al., 2010 ; Koler et al., 2009 ; Li et al., 2007 ; Mirkin et al., 2005 ; Taylor, 2000 ; Taylor et al., 1998 ; Taylor et al., 1997 ; Tulac et al., 2006 ).

Human population growth is a global issue that has become increasingly important since the last half of the twentieth century. By 2020, there will be approximate 1.2 billion people (15% of the world’s population) entering their child-bearing years ( Rao, 2008 ). Meanwhile, about 15% of couples at reproductive age face the risk of infertility. Moreover, human reproduction is an inefficient process with only about 30% success rate in each normal menstrual cycle, while embryo implantation is a major limiting event. Therefore, unraveling the complexities of implantation will help to address these two contrasting global issues ( Wang and Dey, 2006 ). Mapping the molecular landscape of implantation is necessary for fully understanding this process. Although studies in the past few years have established the role of various signaling pathways, whether these pathways function independently or converge into larger networks needs to be further explored. In recent years, the identification of specific genes that regulate implantation has been largely advanced by the PR-Cre mouse model ( Soyal et al., 2005 ), where candidate genes can be ablated in PR-expressing tissues, including the uterus. Employing this mouse model, a number of genes with embryonic lethal phenotypes have been recognized to be essential for adult uterine biology and implantation ( Table 3 ) ( Afshar et al., 2012 ; Daikoku et al., 2008 ; Das, 2012 ; Hawkins et al., 2012 ; Kim et al., 2010a ; Kim et al., 2010b ; Laws et al., 2008 ; Park et al., 2012a ; Park et al., 2012b ). In addition, transgenic or knock-in lines with Cre recombinase driven by the promoter of Wnt7a, small proline-rich protein 2F (Sprr2f) or anti-Müllerian hormone receptor type 2 (AMHR2) known respectively as Wnt7a-Cre, Sprr2f-Cre and AMHR2-Cre mice, have been used for conditional depletion of genes in the uterine epithelium ( Contreras et al., 2010 ; Franco et al., 2011b ; Winuthayanon et al., 2010 ) or in mesenchyme-derived stroma ( Deutscher and Hung-Chang Yao, 2007 ; Jamin et al., 2002 ; Orvis et al., 2008 ). However, currently available Cre mouse lines are not ideal for investigating the potential roles of genes involved in the development of the female reproductive tract during embryo implantation in adulthood, since it will be difficult to distinguish the outcome of any early pregnancy defects from an impaired postnatal uterine development ( Dunlap et al., 2011 ; Jeong et al., 2009 ). Therefore, development of new Cre mouse lines is warranted for conditional gene deletion in epithelial and stromal cells, respectively. To improve the diagnosis and treatment of infertility is one of the ultimate goals of understanding the molecular basis of implantation. Embryo quality and endometrial receptivity are two determinants for pregnancy success in patients undergoing IVF-ET program. With respect to embryos, clinical data show that 70% of embryos selected for transfer fail to implant. To increase pregnancy rates, multiple embryos are transferred, which is responsible for complications due to multiple gestation pregnancy ( Hernandez, 2001 ; Tiitinen, 2012 ). A pre-selection of high quality embryos by monitoring embryonic metabolic activity and morphology can significantly reduce IVF complications, while maintaining acceptable pregnancy rates ( Gardner and Sakkas, 2003 ; Giorgetti et al., 1995 ; Rijnders and Jansen, 1998 ). Newly-developed time-lapse imaging and embryo secretomics technologies can be useful in assessing embryo quality. In addition, improvement of culture medium and embryo freezing techniques are of equal importance ( Ciray et al., 2012 ; Pool et al., 2012 ; Rato et al., 2012 ), particularly since the risk of low birth weight in IVF-ET conceived newborns is greatly reduced in frozen embryo cycles ( Kalra et al., 2011 ). Transferring the embryo into the non-receptive endometrium largely accounts for low pregnancy rates in IVF-ET programs ( Miller et al., 2012 ). Despite numerous advances in our understanding on the underlying molecular and cellular mechanisms that are essential for endometrial receptivity, it has not yet been possible for clinicians to accurately predict when receptivity is attained in each individual. Thus, it is essential to identify reliable biomarkers to define the status of endometrial receptivity. Large-scale technologies, such as proteomics and secretomics, can also be used to screen novel biomarkers for dating endometrium ( Aghajanova et al., 2008b ; Borthwick et al., 2003 ; Carson et al., 2002 ; Diaz-Gimeno et al., 2011 ; Hannan et al., 2010 ; Haouzi et al., 2009 ; Horcajadas et al., 2007 ; Horcajadas et al., 2004 ; Mirkin et al., 2005 ; Riesewijk et al., 2003 ). Characterization of biomarkers of endometrial receptivity will help to formulate new strategies of non-hormonal contraception. Moreover, a better understanding of the window of implantation in humans will suggest new interventions to correct receptivity defects and extend the receptivity window in IVF-ET programs. There is an urgent need to verify in humans about the molecules known to be essential for implantation in mice through expression and functional analysis under various pathophysiological conditions. These efforts will be advanced through large-scale approaches, using mass spectrometry or genome-wide screening to discover essential regulators in humans. Mouse models will continue to be important as functional tools to demonstrate the involvement of embryo-uterine crosstalk during early pregnancy. A better understanding of the physiological and molecular mechanisms responsible for implantation will, thus, help improve modern reproductive therapies and develop novel contraceptives.